[
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.1579051027074456,
    "compile_target": "AUTO:GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394921,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.026748363859951496,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 8.752053,
            "dram": 2.2796570000000003,
            "package-0": 22.433108999999998,
            "psys": 37.295436,
            "uncore": 8.877541
          },
          "net_j": {
            "core": 6.604306042694602,
            "dram": 1.8505633607260248,
            "package-0": 18.67715326585836,
            "psys": 28.611885485606145,
            "uncore": 8.87568399868514
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.818965,
            "dram": 0.127502,
            "package-0": 2.438226,
            "psys": 3.364981,
            "uncore": 0.34082
          },
          "logits": {
            "l2": 401.4866824038882,
            "sha256": "61d2cdecbae6d5fd94941833b71375ae3e9c75ee5788bfeb46933416f5518666",
            "sum": 39969.7027254026,
            "top1": 546
          },
          "net_j": {
            "core": 1.6800812192550942,
            "dram": 0.09975471500757785,
            "package-0": 2.19534760555331,
            "psys": 2.8034603228893857,
            "uncore": 0.34069991724882737
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.0565368,
          "watts": {
            "core": 32.17311556366827,
            "dram": 2.2552036903397434,
            "package-0": 43.126353100989085,
            "psys": 59.51841986104626,
            "uncore": 6.028286001330107
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1171.2164163915008,
          "decode_tokens_per_s": 25.96060966581099,
          "end_to_end_ms": 874.3046850000001,
          "handoff_estimate_ms": 19.738152999999993,
          "handoff_plus_first_decode_ms": 56.5368,
          "output_tokens_per_s_e2e": 18.3002565061172,
          "prefill_tokens_per_s": 3399.591231531468,
          "steady_tpot_ms": 36.798647,
          "steady_tpot_p95_ms": 39.3622204,
          "ttft_ms": 296.506236
        },
        "prefill": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 30,
          "gross_j": {
            "core": 1.622066,
            "dram": 0.579466,
            "package-0": 7.125531,
            "psys": 11.968963,
            "uncore": 4.178822
          },
          "logits": {
            "l2": 628.2510355324671,
            "sha256": "c96d07293a33b871695e60a2f64ee2de15b94f1c3e0c78181714540acc5414c9",
            "sum": -187389.82698735525,
            "top1": 5809
          },
          "net_j": {
            "core": 0.8936924968989884,
            "dram": 0.43394586033305066,
            "package-0": 5.851759606091761,
            "psys": 9.024077856054117,
            "uncore": 4.178192228160088
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.296506236,
          "watts": {
            "core": 5.470596577941788,
            "dram": 1.9543130283438628,
            "package-0": 24.031639590878623,
            "psys": 40.36664847750454,
            "uncore": 14.093538322748802
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 518.3678398516904,
            "sha256": "f563a91e160dc6bc68c1aa9a1c95954b91970ab4b372530c63a6fe21c1c8c083",
            "sum": 92807.67067616689,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 52,
          "gross_j": {
            "core": 5.311022,
            "dram": 1.572689,
            "package-0": 12.869352,
            "psys": 21.961492,
            "uncore": 4.357899
          },
          "net_j": {
            "core": 4.030532326540519,
            "dram": 1.3168627853853963,
            "package-0": 10.63004605421329,
            "psys": 16.784347306662642,
            "uncore": 4.356791853276225
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.521261649,
          "token_latencies_s": [
            0.04130571,
            0.038315726,
            0.037117336,
            0.037018289,
            0.036638301,
            0.036730731,
            0.036755856,
            0.036591956,
            0.036983013,
            0.036872688,
            0.036731281,
            0.036818849,
            0.036778445,
            0.036579627
          ],
          "watts": {
            "core": 10.18878333019278,
            "dram": 3.0170817343211063,
            "package-0": 24.68885256509634,
            "psys": 42.1314171915993,
            "uncore": 8.360290860377491
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.155985,
            "dram": 2.33764,
            "package-0": 22.12891,
            "psys": 37.627772,
            "uncore": 9.049781
          },
          "net_j": {
            "core": 5.891750015920376,
            "dram": 1.8852734741457975,
            "package-0": 18.169241314184383,
            "psys": 28.473249027060497,
            "uncore": 9.04782327992116
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2050,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.551327,
            "dram": 0.122925,
            "package-0": 2.043879,
            "psys": 2.893364,
            "uncore": 0.232544
          },
          "logits": {
            "l2": 401.4866824038882,
            "sha256": "61d2cdecbae6d5fd94941833b71375ae3e9c75ee5788bfeb46933416f5518666",
            "sum": 39969.7027254026,
            "top1": 546
          },
          "net_j": {
            "core": 1.4298406575613827,
            "dram": 0.09865351093013647,
            "package-0": 1.8314250500827818,
            "psys": 2.4021828635135023,
            "uncore": 0.23243895955430008
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.049454652,
          "watts": {
            "core": 31.36867690424755,
            "dram": 2.4856104537951254,
            "package-0": 41.32834662348852,
            "psys": 58.505396014110055,
            "uncore": 4.7021663401857525
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1110.9608818598306,
          "decode_tokens_per_s": 25.240822816698234,
          "end_to_end_ms": 921.7246230000001,
          "handoff_estimate_ms": 11.256723000000003,
          "handoff_plus_first_decode_ms": 49.454652,
          "output_tokens_per_s_e2e": 17.358763779059853,
          "prefill_tokens_per_s": 3078.3398461903835,
          "steady_tpot_ms": 38.197929,
          "steady_tpot_p95_ms": 42.97029485,
          "ttft_ms": 327.449226
        },
        "prefill": {
          "gpu_active_fraction": 0.7272727272727273,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 1.901972,
            "dram": 0.625486,
            "package-0": 7.700236,
            "psys": 12.976041,
            "uncore": 4.367176
          },
          "logits": {
            "l2": 628.2510355324671,
            "sha256": "c96d07293a33b871695e60a2f64ee2de15b94f1c3e0c78181714540acc5414c9",
            "sum": -187389.82698735525,
            "top1": 5809
          },
          "net_j": {
            "core": 1.0975864226614687,
            "dram": 0.4647795752577073,
            "package-0": 6.293535545186653,
            "psys": 9.723831623047879,
            "uncore": 4.366480506022141
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.327449226,
          "watts": {
            "core": 5.8084486050976345,
            "dram": 1.9101770605498394,
            "package-0": 23.515816769711957,
            "psys": 39.627642912797725,
            "uncore": 13.336956246157076
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 518.3678398516904,
            "sha256": "f563a91e160dc6bc68c1aa9a1c95954b91970ab4b372530c63a6fe21c1c8c083",
            "sum": 92807.67067616689,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 54,
          "gross_j": {
            "core": 4.702686,
            "dram": 1.589229,
            "package-0": 12.384795,
            "psys": 21.758367,
            "uncore": 4.450061
          },
          "net_j": {
            "core": 3.3643229356975244,
            "dram": 1.3218403879579537,
            "package-0": 10.044280718914948,
            "psys": 16.347234540499116,
            "uncore": 4.448903814344718
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.544820745,
          "token_latencies_s": [
            0.04185664,
            0.045038511,
            0.038284758,
            0.038300555,
            0.038141586,
            0.038000403,
            0.03829248,
            0.038045776,
            0.038221264,
            0.038348972,
            0.037997244,
            0.038174594,
            0.038014085,
            0.038076784
          ],
          "watts": {
            "core": 8.631620662682366,
            "dram": 2.916975931230372,
            "package-0": 22.73187119554341,
            "psys": 39.9367446993965,
            "uncore": 8.167936042890585
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.685586,
            "dram": 2.333246,
            "package-0": 22.631656,
            "psys": 38.226282,
            "uncore": 9.028542
          },
          "net_j": {
            "core": 6.4339668060840545,
            "dram": 1.8833999551061995,
            "package-0": 18.694049666196953,
            "psys": 29.122765897675617,
            "uncore": 9.026595187881634
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.628902,
            "dram": 0.12439,
            "package-0": 2.120417,
            "psys": 3.002007,
            "uncore": 0.227905
          },
          "logits": {
            "l2": 401.4866824038882,
            "sha256": "61d2cdecbae6d5fd94941833b71375ae3e9c75ee5788bfeb46933416f5518666",
            "sum": 39969.7027254026,
            "top1": 546
          },
          "net_j": {
            "core": 1.503730968431162,
            "dram": 0.09938235499561084,
            "package-0": 1.9015193071548229,
            "psys": 2.495928305827166,
            "uncore": 0.2277967736703502
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.050954615,
          "watts": {
            "core": 31.967703023563224,
            "dram": 2.4411920294167664,
            "package-0": 41.613836155959575,
            "psys": 58.915311203901744,
            "uncore": 4.472705759821755
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1117.1855798031909,
          "decode_tokens_per_s": 25.393661298258756,
          "end_to_end_ms": 916.588988,
          "handoff_estimate_ms": 12.7774825,
          "handoff_plus_first_decode_ms": 50.954615000000004,
          "output_tokens_per_s_e2e": 17.456024684424857,
          "prefill_tokens_per_s": 3093.0644642042917,
          "steady_tpot_ms": 38.1771325,
          "steady_tpot_p95_ms": 40.4363914,
          "ttft_ms": 325.890395
        },
        "prefill": {
          "gpu_active_fraction": 0.7272727272727273,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 2.134638,
            "dram": 0.620421,
            "package-0": 7.909281,
            "psys": 13.265957,
            "uncore": 4.354664
          },
          "logits": {
            "l2": 628.2510355324671,
            "sha256": "c96d07293a33b871695e60a2f64ee2de15b94f1c3e0c78181714540acc5414c9",
            "sum": -187389.82698735525,
            "top1": 5809
          },
          "net_j": {
            "core": 1.3340817222118915,
            "dram": 0.4604796225899232,
            "package-0": 6.509277181191763,
            "psys": 10.029229854773428,
            "uncore": 4.353971816940252
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.325890395,
          "watts": {
            "core": 6.550171569186627,
            "dram": 1.903771972168741,
            "package-0": 24.26975793502598,
            "psys": 40.70680573448628,
            "uncore": 13.362357611061228
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 518.3678398516904,
            "sha256": "f563a91e160dc6bc68c1aa9a1c95954b91970ab4b372530c63a6fe21c1c8c083",
            "sum": 92807.67067616689,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 53,
          "gross_j": {
            "core": 4.922046,
            "dram": 1.588435,
            "package-0": 12.601958,
            "psys": 21.958318,
            "uncore": 4.445973
          },
          "net_j": {
            "core": 3.596154115441001,
            "dram": 1.3235379775206655,
            "package-0": 10.283253177850366,
            "psys": 16.597607737075023,
            "uncore": 4.444826597271033
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.539743978,
          "token_latencies_s": [
            0.042332392,
            0.039415468,
            0.038536746,
            0.038111865,
            0.038129457,
            0.038193244,
            0.038175865,
            0.038246715,
            0.038066327,
            0.038120735,
            0.038244376,
            0.0381784,
            0.038130421,
            0.037837393
          ],
          "watts": {
            "core": 9.119223558988926,
            "dram": 2.9429415885025403,
            "package-0": 23.348028905660158,
            "psys": 40.68284018909424,
            "uncore": 8.237188706531526
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.0791938668116927,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394990,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.026207742281258106,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 8.934059,
            "dram": 2.517695,
            "package-0": 20.726571,
            "psys": 36.344755,
            "uncore": 4.616504
          },
          "net_j": {
            "core": 6.246924785246469,
            "dram": 1.9808383672834422,
            "package-0": 16.027340893780103,
            "psys": 25.48040917313246,
            "uncore": 4.615354077636267
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 17,
          "gross_j": {
            "core": 3.006828,
            "dram": 0.215576,
            "package-0": 3.596793,
            "psys": 5.600449,
            "uncore": 0.00531
          },
          "logits": {
            "l2": 403.29190860783467,
            "sha256": "1dfee3a1ff1e7c218e8d6915056b5ffa93b645fba18d446b92fe98015d2630a1",
            "sum": 41609.541811217554,
            "top1": 546
          },
          "net_j": {
            "core": 2.5687796953541797,
            "dram": 0.12805929297875357,
            "package-0": 2.8307390493382787,
            "psys": 3.8293769618354565,
            "uncore": 0.0049312513420483005
          },
          "npu_busy_delta_us": 52717,
          "npu_utilization": 0.29563033167988134,
          "seconds": 0.178320674,
          "watts": {
            "core": 16.861914732332156,
            "dram": 1.2089232009071476,
            "package-0": 20.17036454225156,
            "psys": 31.406616374722763,
            "uncore": 0.029777814769811825
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 936.1186653200733,
          "decode_tokens_per_s": 20.525507423692204,
          "end_to_end_ms": 1093.8784130000001,
          "handoff_estimate_ms": 140.52945250000002,
          "handoff_plus_first_decode_ms": 178.32067400000003,
          "output_tokens_per_s_e2e": 14.626854145626146,
          "prefill_tokens_per_s": 2776.2445850767785,
          "steady_tpot_ms": 37.7912215,
          "steady_tpot_p95_ms": 46.6162753,
          "ttft_ms": 363.08040200000005
        },
        "prefill": {
          "gpu_active_fraction": 0.9722222222222222,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 2.232966,
            "dram": 0.675841,
            "package-0": 8.485574,
            "psys": 14.76516,
            "uncore": 4.611194
          },
          "logits": {
            "l2": 628.3565558181504,
            "sha256": "d9e1e7e1f931cd6721a36966da9356a1de9f71a7b58989f1fd1d69ea34d47bf0",
            "sum": -187423.86018045805,
            "top1": 5809
          },
          "net_j": {
            "core": 1.3410517256841517,
            "dram": 0.4976474094632663,
            "package-0": 6.925804343903108,
            "psys": 11.159063560572331,
            "uncore": 4.610422826294219
          },
          "npu_busy_delta_us": 9371,
          "npu_utilization": 0.025809710324161204,
          "seconds": 0.363080402,
          "watts": {
            "core": 6.150059291825945,
            "dram": 1.8614086474433285,
            "package-0": 23.371060385682835,
            "psys": 40.6663645811431,
            "uncore": 12.700200767101718
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 513.951448933101,
            "sha256": "6251fce17c6ac65a0e5a8579de91fe30d8b29d6fafc4fb59eee5beae93676e4d",
            "sum": 91180.61728955619,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 3.694265,
            "dram": 1.626278,
            "package-0": 8.644204,
            "psys": 15.979146,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.337093364208137,
            "dram": 1.3551316648414224,
            "package-0": 6.270797500538715,
            "psys": 10.491968650724674,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 541473,
          "npu_utilization": 0.9800818309403341,
          "seconds": 0.552477337,
          "token_latencies_s": [
            0.049497386,
            0.045064908,
            0.040802834,
            0.03939486,
            0.038527777,
            0.037805866,
            0.037604214,
            0.037743968,
            0.037820727,
            0.037722436,
            0.03753061,
            0.037776577,
            0.037596288,
            0.037560723
          ],
          "watts": {
            "core": 6.686726771563483,
            "dram": 2.9436103367259023,
            "package-0": 15.646259893552882,
            "psys": 28.92271760280368,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 7.549419,
            "dram": 2.438226,
            "package-0": 19.059827,
            "psys": 33.292578,
            "uncore": 4.604053
          },
          "net_j": {
            "core": 5.158998414563909,
            "dram": 1.960649130475725,
            "package-0": 14.879486353100308,
            "psys": 23.627874328239827,
            "uncore": 4.603285310161245
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.16666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.763668,
            "dram": 0.132019,
            "package-0": 2.054377,
            "psys": 2.92498,
            "uncore": 0.0
          },
          "logits": {
            "l2": 403.29190860783467,
            "sha256": "1dfee3a1ff1e7c218e8d6915056b5ffa93b645fba18d446b92fe98015d2630a1",
            "sum": 41609.541811217554,
            "top1": 546
          },
          "net_j": {
            "core": 1.626144816908459,
            "dram": 0.10454354581924957,
            "package-0": 1.8138780034337878,
            "psys": 2.3689603519620004,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 50833,
          "npu_utilization": 0.908008955873119,
          "seconds": 0.055982928,
          "watts": {
            "core": 31.503675549088822,
            "dram": 2.358201057293752,
            "package-0": 36.69649075875417,
            "psys": 52.247713803036525,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1052.3154418836605,
          "decode_tokens_per_s": 24.523744588437246,
          "end_to_end_ms": 973.092249,
          "handoff_estimate_ms": 17.964838499999995,
          "handoff_plus_first_decode_ms": 55.982928,
          "output_tokens_per_s_e2e": 16.442428779432195,
          "prefill_tokens_per_s": 2788.843510420912,
          "steady_tpot_ms": 38.0180895,
          "steady_tpot_p95_ms": 46.445875050000005,
          "ttft_ms": 361.44014400000003
        },
        "prefill": {
          "gpu_active_fraction": 0.9722222222222222,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 1.973444,
            "dram": 0.673033,
            "package-0": 8.240091,
            "psys": 14.31344,
            "uncore": 4.604053
          },
          "logits": {
            "l2": 628.3565558181504,
            "sha256": "d9e1e7e1f931cd6721a36966da9356a1de9f71a7b58989f1fd1d69ea34d47bf0",
            "sum": -187423.86018045805,
            "top1": 5809
          },
          "net_j": {
            "core": 1.08555905228978,
            "dram": 0.4956444197607557,
            "package-0": 6.687367784973674,
            "psys": 10.723634521199678,
            "uncore": 4.603285310161245
          },
          "npu_busy_delta_us": 8154,
          "npu_utilization": 0.022559751968226304,
          "seconds": 0.361440144,
          "watts": {
            "core": 5.459946917241157,
            "dram": 1.8620870182034897,
            "package-0": 22.79794078435294,
            "psys": 39.601135174403865,
            "uncore": 12.73807870107533
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 513.951448933101,
            "sha256": "6251fce17c6ac65a0e5a8579de91fe30d8b29d6fafc4fb59eee5beae93676e4d",
            "sum": 91180.61728955619,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 3.812307,
            "dram": 1.633174,
            "package-0": 8.765359,
            "psys": 16.054158,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.44729454536567,
            "dram": 1.3604611648957197,
            "package-0": 6.378240564692846,
            "psys": 10.535279455078147,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 543663,
          "npu_utilization": 0.9783933003719586,
          "seconds": 0.555669177,
          "token_latencies_s": [
            0.050866189,
            0.044065706,
            0.040956959,
            0.039648773,
            0.038754758,
            0.038050713,
            0.037901184,
            0.037985466,
            0.037876823,
            0.037906122,
            0.037800944,
            0.0378264,
            0.037925677,
            0.038080741
          ],
          "watts": {
            "core": 6.860749449127715,
            "dram": 2.9391120969086972,
            "package-0": 15.77442003769808,
            "psys": 28.891575535419705,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 7.59489,
            "dram": 2.436823,
            "package-0": 19.117382,
            "psys": 32.888831,
            "uncore": 4.608082
          },
          "net_j": {
            "core": 5.2074352153945425,
            "dram": 1.9598386612985248,
            "package-0": 14.942227912251631,
            "psys": 23.236118350385535,
            "uncore": 4.60731365773134
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.781306,
            "dram": 0.134277,
            "package-0": 2.080683,
            "psys": 2.952568,
            "uncore": 0.0
          },
          "logits": {
            "l2": 403.29190860783467,
            "sha256": "1dfee3a1ff1e7c218e8d6915056b5ffa93b645fba18d446b92fe98015d2630a1",
            "sum": 41609.541811217554,
            "top1": 546
          },
          "net_j": {
            "core": 1.6432835144764308,
            "dram": 0.10670179128488577,
            "package-0": 1.8393108289537736,
            "psys": 2.394529623509969,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 51225,
          "npu_utilization": 0.911700997526367,
          "seconds": 0.056186184,
          "watts": {
            "core": 31.70363020204398,
            "dram": 2.3898579764733623,
            "package-0": 37.031932974839506,
            "psys": 52.549715780662375,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1053.6226741846665,
          "decode_tokens_per_s": 24.58461776556627,
          "end_to_end_ms": 971.8849309999999,
          "handoff_estimate_ms": 18.176638500000003,
          "handoff_plus_first_decode_ms": 56.186184,
          "output_tokens_per_s_e2e": 16.462854284135414,
          "prefill_tokens_per_s": 2786.4753927491456,
          "steady_tpot_ms": 38.0095455,
          "steady_tpot_p95_ms": 45.8224811,
          "ttft_ms": 361.747318
        },
        "prefill": {
          "gpu_active_fraction": 0.9722222222222222,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 2.171137,
            "dram": 0.674376,
            "package-0": 8.437417,
            "psys": 14.375024,
            "uncore": 4.608082
          },
          "logits": {
            "l2": 628.3565558181504,
            "sha256": "d9e1e7e1f931cd6721a36966da9356a1de9f71a7b58989f1fd1d69ea34d47bf0",
            "sum": -187423.86018045805,
            "top1": 5809
          },
          "net_j": {
            "core": 1.2824974732109051,
            "dram": 0.49683666406676064,
            "package-0": 6.883374185563453,
            "psys": 10.782167684355873,
            "uncore": 4.60731365773134
          },
          "npu_busy_delta_us": 8071,
          "npu_utilization": 0.02231115366555392,
          "seconds": 0.361747318,
          "watts": {
            "core": 6.001805381733335,
            "dram": 1.8642183824013867,
            "package-0": 23.324062350062814,
            "psys": 39.73774865692301,
            "uncore": 12.738399901557804
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 513.951448933101,
            "sha256": "6251fce17c6ac65a0e5a8579de91fe30d8b29d6fafc4fb59eee5beae93676e4d",
            "sum": 91180.61728955619,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 3.642447,
            "dram": 1.62817,
            "package-0": 8.599282,
            "psys": 15.561239,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.2816542277072065,
            "dram": 1.3563002059468785,
            "package-0": 6.219542897734405,
            "psys": 10.059421042519695,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 543178,
          "npu_utilization": 0.9805516721575241,
          "seconds": 0.553951429,
          "token_latencies_s": [
            0.050091308,
            0.043523882,
            0.041000158,
            0.039627643,
            0.039086622,
            0.038120256,
            0.038038178,
            0.037786302,
            0.037980913,
            0.037874564,
            0.037727016,
            0.037719199,
            0.037585589,
            0.037760515
          ],
          "watts": {
            "core": 6.5753905655147245,
            "dram": 2.9391927067309718,
            "package-0": 15.523530674022325,
            "psys": 28.091341921603743,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 14.522015229798853,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395128,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU+GPU",
      "key": "npu_gpu_combined",
      "label": "NPU + iGPU (subgraph combination)",
      "mode": "combined",
      "prefill_device": "NPU+GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.026272007264196873,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 47.860779,
            "dram": 8.310891,
            "package-0": 102.183393,
            "psys": 183.774858,
            "uncore": 28.635181
          },
          "net_j": {
            "core": 33.41361216364959,
            "dram": 5.417603169704745,
            "package-0": 76.85777634647256,
            "psys": 125.22350993163995,
            "uncore": 28.62265963213676
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.40404040404040403,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 99,
          "gross_j": {
            "core": 12.692472,
            "dram": 0.972531,
            "package-0": 17.322466,
            "psys": 28.4229,
            "uncore": 1.863399
          },
          "logits": {
            "l2": 430.57070801613617,
            "sha256": "0cacbe9345c91be7436afb7ddea70f4d375bcbddb0707a37cae2ac55ada07d2d",
            "sum": -86575.56225838233,
            "top1": 90
          },
          "net_j": {
            "core": 10.24428880786052,
            "dram": 0.48341386636779454,
            "package-0": 13.041110755859878,
            "psys": 18.524656473162462,
            "uncore": 1.861282233221979
          },
          "npu_busy_delta_us": 10004,
          "npu_utilization": 0.01003806671904195,
          "seconds": 0.996606247,
          "watts": {
            "core": 12.735693799037566,
            "dram": 0.9758427693259282,
            "package-0": 17.381454362888416,
            "psys": 28.519688779354,
            "uncore": 1.869744450839269
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 173.6991077156562,
          "decode_tokens_per_s": 2.73146209932149,
          "end_to_end_ms": 5895.251929999999,
          "handoff_estimate_ms": 677.2658710000001,
          "handoff_plus_first_decode_ms": 996.606247,
          "output_tokens_per_s_e2e": 2.714048558057128,
          "prefill_tokens_per_s": 2496.9807155726944,
          "steady_tpot_ms": 319.340376,
          "steady_tpot_p95_ms": 332.3457439,
          "ttft_ms": 403.68753899999996
        },
        "prefill": {
          "gpu_active_fraction": 0.1,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 40,
          "gross_j": {
            "core": 0.957029,
            "dram": 0.582091,
            "package-0": 5.999923,
            "psys": 10.481174,
            "uncore": 0.234924
          },
          "logits": {
            "l2": 633.1706927261538,
            "sha256": "ad53eb322905ce268ba906538fffea0ad7f5b54a7f40119f6d9c48fbf51f50d8",
            "sum": -190262.31765024643,
            "top1": 5809
          },
          "net_j": {
            "core": 0.0,
            "dram": 0.3839681279498922,
            "package-0": 4.265707739665926,
            "psys": 6.4717695016839,
            "uncore": 0.23406657774940848
          },
          "npu_busy_delta_us": 347381,
          "npu_utilization": 0.8605195019408315,
          "seconds": 0.403687539,
          "watts": {
            "core": 2.370717219487917,
            "dram": 1.4419345255043903,
            "package-0": 14.862789708255027,
            "psys": 25.96358070889079,
            "uncore": 0.5819451365329362
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 569.9173451531581,
            "sha256": "a40d57a3dc57a3450a26a1b67cdc6755d927ed0f29960e651636fd033560248b",
            "sum": 139928.31610714644,
            "top1": 802
          },
          "gpu_active_fraction": 0.9977528089887641,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 445,
          "gross_j": {
            "core": 34.211278,
            "dram": 6.756269,
            "package-0": 78.861004,
            "psys": 144.870784,
            "uncore": 26.536858
          },
          "net_j": {
            "core": 23.16932335578907,
            "dram": 4.550221175387058,
            "package-0": 59.55095785094676,
            "psys": 100.22708395679359,
            "uncore": 26.527310821165372
          },
          "npu_busy_delta_us": 137087,
          "npu_utilization": 0.030497948058312328,
          "seconds": 4.494958144,
          "token_latencies_s": [
            0.337700947,
            0.31774869,
            0.317453958,
            0.318008498,
            0.319460646,
            0.319301666,
            0.318089272,
            0.318683853,
            0.319202785,
            0.319379086,
            0.319390041,
            0.319846872,
            0.329462173,
            0.321176748
          ],
          "watts": {
            "core": 7.6110337191162065,
            "dram": 1.5030771774857266,
            "package-0": 17.544324434981878,
            "psys": 32.2296180206656,
            "uncore": 5.903694127924676
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 42.552443,
            "dram": 8.022929,
            "package-0": 96.140501,
            "psys": 170.289236,
            "uncore": 28.966662
          },
          "net_j": {
            "core": 29.467158532994233,
            "dram": 5.408648805685504,
            "package-0": 73.2571026277884,
            "psys": 117.38415410570833,
            "uncore": 28.955348101787788
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9722222222222222,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 4.382435,
            "dram": 0.513548,
            "package-0": 7.701823,
            "psys": 12.754606,
            "uncore": 1.913691
          },
          "logits": {
            "l2": 634.5086714166829,
            "sha256": "1fbf009fd116a1af69f35aa003ae823419111749aab8664994f7bf095dbdd55c",
            "sum": 167224.29535845853,
            "top1": 802
          },
          "net_j": {
            "core": 3.501240428766107,
            "dram": 0.337496075357252,
            "package-0": 6.160799840142815,
            "psys": 9.191850366105704,
            "uncore": 1.9129290948454554
          },
          "npu_busy_delta_us": 8042,
          "npu_utilization": 0.022418810328573957,
          "seconds": 0.358716626,
          "watts": {
            "core": 12.216983218391444,
            "dram": 1.4316258650358737,
            "package-0": 21.470493536588965,
            "psys": 35.5562164548236,
            "uncore": 5.334826604886723
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 192.2370507873706,
          "decode_tokens_per_s": 3.0536580641770157,
          "end_to_end_ms": 5326.756708999999,
          "handoff_estimate_ms": 34.05667150000002,
          "handoff_plus_first_decode_ms": 358.71662599999996,
          "output_tokens_per_s_e2e": 3.0037039185526657,
          "prefill_tokens_per_s": 2431.168893413972,
          "steady_tpot_ms": 324.65995449999997,
          "steady_tpot_p95_ms": 331.10038959999997,
          "ttft_ms": 414.615374
        },
        "prefill": {
          "gpu_active_fraction": 0.0975609756097561,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 41,
          "gross_j": {
            "core": 1.672237,
            "dram": 0.597777,
            "package-0": 6.67552,
            "psys": 10.899813,
            "uncore": 0.236938
          },
          "logits": {
            "l2": 633.1706927261538,
            "sha256": "ad53eb322905ce268ba906538fffea0ad7f5b54a7f40119f6d9c48fbf51f50d8",
            "sum": -190262.31765024643,
            "top1": 5809
          },
          "net_j": {
            "core": 0.6537260355293704,
            "dram": 0.39429093526662806,
            "package-0": 4.89435947534997,
            "psys": 6.7818737896281585,
            "uncore": 0.23605736730079518
          },
          "npu_busy_delta_us": 348123,
          "npu_utilization": 0.8396287784543176,
          "seconds": 0.414615374,
          "watts": {
            "core": 4.033224778587202,
            "dram": 1.4417627456332578,
            "package-0": 16.100512471590115,
            "psys": 26.288974513521055,
            "uncore": 0.5714645786386108
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1087.0447466317994,
            "sha256": "1ffb883bcf5cae00e56811d75cd2b9c78794a471985aba335602ec78b832ac9a",
            "sum": 326622.44280824065,
            "top1": 8940
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 449,
          "gross_j": {
            "core": 36.497771,
            "dram": 6.911604,
            "package-0": 81.763158,
            "psys": 146.634817,
            "uncore": 26.816033
          },
          "net_j": {
            "core": 25.312192068698756,
            "dram": 4.676861795061624,
            "package-0": 62.20194331229561,
            "psys": 101.41042994997447,
            "uncore": 26.806361639641537
          },
          "npu_busy_delta_us": 136077,
          "npu_utilization": 0.029884539373416927,
          "seconds": 4.553424709,
          "token_latencies_s": [
            0.334018198,
            0.321224467,
            0.32096755,
            0.321251807,
            0.321796775,
            0.325829801,
            0.326582775,
            0.326650077,
            0.329529262,
            0.328230902,
            0.328368279,
            0.322585459,
            0.322820395,
            0.323490108
          ],
          "watts": {
            "core": 8.015455032749506,
            "dram": 1.5178913546849644,
            "package-0": 17.956409345781502,
            "psys": 32.20319350184297,
            "uncore": 5.889200923207799
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 44.61212999999999,
            "dram": 8.165994,
            "package-0": 98.58049700000001,
            "psys": 173.889752,
            "uncore": 28.972948000000002
          },
          "net_j": {
            "core": 31.391813381566674,
            "dram": 5.524736029240304,
            "package-0": 75.46095591939219,
            "psys": 120.43872194174978,
            "uncore": 28.961517349261662
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.972972972972973,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 37,
          "gross_j": {
            "core": 4.788135,
            "dram": 0.538634,
            "package-0": 8.207682,
            "psys": 13.459682,
            "uncore": 1.955073
          },
          "logits": {
            "l2": 1085.5057909602467,
            "sha256": "eda78e61e9193b4d2f161191c2f713a7b36c6702e1f3ac7fda9b3c8bd416522c",
            "sum": 325965.88127466664,
            "top1": 8940
          },
          "net_j": {
            "core": 3.876696525697236,
            "dram": 0.3565397125036909,
            "package-0": 6.613768642455436,
            "psys": 9.774647314955068,
            "uncore": 1.9542849451318804
          },
          "npu_busy_delta_us": 7865,
          "npu_utilization": 0.02119784325245778,
          "seconds": 0.371028312,
          "watts": {
            "core": 12.905039440763753,
            "dram": 1.4517328801582126,
            "package-0": 22.121443929055204,
            "psys": 36.27669793565511,
            "uncore": 5.26933642735059
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 190.27354391373697,
          "decode_tokens_per_s": 3.0181301006750814,
          "end_to_end_ms": 5381.725588,
          "handoff_estimate_ms": 46.12672950000002,
          "handoff_plus_first_decode_ms": 371.028312,
          "output_tokens_per_s_e2e": 2.97302412365214,
          "prefill_tokens_per_s": 2448.0225791614002,
          "steady_tpot_ms": 324.9015825,
          "steady_tpot_p95_ms": 343.38017314999996,
          "ttft_ms": 411.760908
        },
        "prefill": {
          "gpu_active_fraction": 0.0975609756097561,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 41,
          "gross_j": {
            "core": 2.226191,
            "dram": 0.594908,
            "package-0": 7.315777,
            "psys": 11.5029,
            "uncore": 0.315917
          },
          "logits": {
            "l2": 633.1706927261538,
            "sha256": "ad53eb322905ce268ba906538fffea0ad7f5b54a7f40119f6d9c48fbf51f50d8",
            "sum": -190262.31765024643,
            "top1": 5809
          },
          "net_j": {
            "core": 1.214692088388531,
            "dram": 0.3928228578799203,
            "package-0": 5.54687907456115,
            "psys": 7.413311203537493,
            "uncore": 0.3150424301152879
          },
          "npu_busy_delta_us": 351326,
          "npu_utilization": 0.8532281554032323,
          "seconds": 0.411760908,
          "watts": {
            "core": 5.406513723736008,
            "dram": 1.4447898973449904,
            "package-0": 17.767050873124653,
            "psys": 27.9358719502338,
            "uncore": 0.7672340765287025
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 944.2549601773937,
            "sha256": "466639dd7dc3f26c0c878b29f045f1c72f6ae69ea88632a94899e0eb977e183b",
            "sum": 267898.8189940732,
            "top1": 3765
          },
          "gpu_active_fraction": 0.9978021978021978,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 455,
          "gross_j": {
            "core": 37.597804,
            "dram": 7.032452,
            "package-0": 83.057038,
            "psys": 148.92717,
            "uncore": 26.701958
          },
          "net_j": {
            "core": 26.30042476748091,
            "dram": 4.775373458856693,
            "package-0": 63.300308202375604,
            "psys": 103.25076342325721,
            "uncore": 26.692189974014493
          },
          "npu_busy_delta_us": 136120,
          "npu_utilization": 0.029598148160331315,
          "seconds": 4.598936368,
          "token_latencies_s": [
            0.361038547,
            0.324808595,
            0.332903266,
            0.324700847,
            0.32499457,
            0.324450015,
            0.325823728,
            0.325840706,
            0.333871818,
            0.323532067,
            0.323924733,
            0.324029129,
            0.323600632,
            0.325364959
          ],
          "watts": {
            "core": 8.175325986593426,
            "dram": 1.5291474891743924,
            "package-0": 18.06005375023706,
            "psys": 32.382959467814054,
            "uncore": 5.8061159936449025
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 4.770954256877303,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394836,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.028868702240288258,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 8.555276000000001,
            "dram": 2.280024,
            "package-0": 22.186527,
            "psys": 37.365445,
            "uncore": 8.826454
          },
          "net_j": {
            "core": 6.38960574697493,
            "dram": 1.8473495049078017,
            "package-0": 18.399227207796013,
            "psys": 28.60942885214928,
            "uncore": 8.82458150170883
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.774104,
            "dram": 0.126953,
            "package-0": 2.380609,
            "psys": 3.3153,
            "uncore": 0.337157
          },
          "logits": {
            "l2": 401.4866824038882,
            "sha256": "61d2cdecbae6d5fd94941833b71375ae3e9c75ee5788bfeb46933416f5518666",
            "sum": 39969.7027254026,
            "top1": 546
          },
          "net_j": {
            "core": 1.639020976499471,
            "dram": 0.09996505993526558,
            "package-0": 2.1443773270944773,
            "psys": 2.7691461552865793,
            "uncore": 0.33704020348847313
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.054989588,
          "watts": {
            "core": 32.26254395650318,
            "dram": 2.3086734165020477,
            "package-0": 43.29199556832469,
            "psys": 60.28959518663788,
            "uncore": 6.131287981281111
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1161.5233164593224,
          "decode_tokens_per_s": 25.521630132804763,
          "end_to_end_ms": 881.6008989999999,
          "handoff_estimate_ms": 17.576560999999998,
          "handoff_plus_first_decode_ms": 54.989588,
          "output_tokens_per_s_e2e": 18.148801819676912,
          "prefill_tokens_per_s": 3430.156452532469,
          "steady_tpot_ms": 37.413027,
          "steady_tpot_p95_ms": 40.821983800000005,
          "ttft_ms": 293.864147
        },
        "prefill": {
          "gpu_active_fraction": 0.8275862068965517,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 1.728512,
            "dram": 0.571471,
            "package-0": 7.179242,
            "psys": 12.130828,
            "uncore": 4.134572
          },
          "logits": {
            "l2": 628.2510355324671,
            "sha256": "c96d07293a33b871695e60a2f64ee2de15b94f1c3e0c78181714540acc5414c9",
            "sum": -187389.82698735525,
            "top1": 5809
          },
          "net_j": {
            "core": 1.0066288414070703,
            "dram": 0.4272475519827078,
            "package-0": 5.916820847603578,
            "psys": 9.212183952194442,
            "uncore": 4.133947839891137
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.293864147,
          "watts": {
            "core": 5.88201050603155,
            "dram": 1.9446775179416491,
            "package-0": 24.430479435111216,
            "psys": 41.280394780517405,
            "uncore": 14.069671452638964
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 518.3678398516904,
            "sha256": "f563a91e160dc6bc68c1aa9a1c95954b91970ab4b372530c63a6fe21c1c8c083",
            "sum": 92807.67067616689,
            "top1": 42
          },
          "gpu_active_fraction": 0.9811320754716981,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 53,
          "gross_j": {
            "core": 5.05266,
            "dram": 1.5816,
            "package-0": 12.626676,
            "psys": 21.919317,
            "uncore": 4.354725
          },
          "net_j": {
            "core": 3.743955929068389,
            "dram": 1.3201368929898283,
            "package-0": 10.338029033097959,
            "psys": 16.628098744668257,
            "uncore": 4.35359345832922
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.532747164,
          "token_latencies_s": [
            0.043741516,
            0.038761713,
            0.037647569,
            0.037422258,
            0.037285907,
            0.037350122,
            0.037403796,
            0.03716556,
            0.037422477,
            0.037353281,
            0.037133919,
            0.037311275,
            0.037469203,
            0.039249928
          ],
          "watts": {
            "core": 9.484161233376364,
            "dram": 2.968762870786488,
            "package-0": 23.701066571984605,
            "psys": 41.14393934155227,
            "uncore": 8.174093255238802
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.692056,
            "dram": 2.336909,
            "package-0": 22.539309,
            "psys": 38.068506,
            "uncore": 8.915444
          },
          "net_j": {
            "core": 6.447430771586349,
            "dram": 1.8884602640862582,
            "package-0": 18.613933634441782,
            "psys": 28.993267182636274,
            "uncore": 8.9135032350573
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2250,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.521297,
            "dram": 0.121155,
            "package-0": 2.035578,
            "psys": 2.853935,
            "uncore": 0.256774
          },
          "logits": {
            "l2": 401.4866824038882,
            "sha256": "61d2cdecbae6d5fd94941833b71375ae3e9c75ee5788bfeb46933416f5518666",
            "sum": 39969.7027254026,
            "top1": 546
          },
          "net_j": {
            "core": 1.4026985031858825,
            "dram": 0.0974604672538147,
            "package-0": 1.828174282045985,
            "psys": 2.3744297051543524,
            "uncore": 0.25667145646535544
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.048279068,
          "watts": {
            "core": 31.5104881477828,
            "dram": 2.5094726352215413,
            "package-0": 42.162744318096614,
            "psys": 59.11329937023639,
            "uncore": 5.318536803568785
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1120.6665873700208,
          "decode_tokens_per_s": 25.2723554572565,
          "end_to_end_ms": 913.7418850000001,
          "handoff_estimate_ms": 10.164394,
          "handoff_plus_first_decode_ms": 48.279068,
          "output_tokens_per_s_e2e": 17.510415427656575,
          "prefill_tokens_per_s": 3147.9540954465865,
          "steady_tpot_ms": 38.114674,
          "steady_tpot_p95_ms": 42.4972871,
          "ttft_ms": 320.207973
        },
        "prefill": {
          "gpu_active_fraction": 0.78125,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 32,
          "gross_j": {
            "core": 2.14355,
            "dram": 0.625548,
            "package-0": 7.826579,
            "psys": 13.158414,
            "uncore": 4.251515
          },
          "logits": {
            "l2": 628.2510355324671,
            "sha256": "c96d07293a33b871695e60a2f64ee2de15b94f1c3e0c78181714540acc5414c9",
            "sum": -187389.82698735525,
            "top1": 5809
          },
          "net_j": {
            "core": 1.3569527055846333,
            "dram": 0.4683954571663621,
            "package-0": 6.450986494388275,
            "psys": 9.978124386594619,
            "uncore": 4.25083488626266
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.320207973,
          "watts": {
            "core": 6.694243056839812,
            "dram": 1.953567845732561,
            "package-0": 24.44217402419271,
            "psys": 41.09333654849375,
            "uncore": 13.277355214387496
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 518.3678398516904,
            "sha256": "f563a91e160dc6bc68c1aa9a1c95954b91970ab4b372530c63a6fe21c1c8c083",
            "sum": 92807.67067616689,
            "top1": 42
          },
          "gpu_active_fraction": 0.9814814814814815,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 54,
          "gross_j": {
            "core": 5.027209,
            "dram": 1.590206,
            "package-0": 12.677152,
            "psys": 22.056157,
            "uncore": 4.407155
          },
          "net_j": {
            "core": 3.687779562815834,
            "dram": 1.3226043396660814,
            "package-0": 10.334772858007522,
            "psys": 16.640713090887303,
            "uncore": 4.405996892329284
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.545254844,
          "token_latencies_s": [
            0.048495088,
            0.039267702,
            0.038694144,
            0.038090107,
            0.037870312,
            0.038040618,
            0.037815981,
            0.037915328,
            0.038186964,
            0.038660971,
            0.038083697,
            0.038139241,
            0.037812673,
            0.038164603
          ],
          "watts": {
            "core": 9.219925426283787,
            "dram": 2.9164454337245647,
            "package-0": 23.249957592307055,
            "psys": 40.45109776227866,
            "uncore": 8.08274341529738
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.220743,
            "dram": 2.34588,
            "package-0": 22.060674,
            "psys": 37.247219,
            "uncore": 8.915016000000001
          },
          "net_j": {
            "core": 5.984695352368028,
            "dram": 1.8991449600788228,
            "package-0": 18.150299011284325,
            "psys": 28.20666017869245,
            "uncore": 8.913082651470456
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2300,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.503842,
            "dram": 0.121887,
            "package-0": 2.022394,
            "psys": 2.825432,
            "uncore": 0.262389
          },
          "logits": {
            "l2": 401.4866824038882,
            "sha256": "61d2cdecbae6d5fd94941833b71375ae3e9c75ee5788bfeb46933416f5518666",
            "sum": 39969.7027254026,
            "top1": 546
          },
          "net_j": {
            "core": 1.3853916534514252,
            "dram": 0.09822206586941742,
            "package-0": 1.8152493655648732,
            "psys": 2.346525691133677,
            "uncore": 0.2622865845601652
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.048218759,
          "watts": {
            "core": 31.187903446457423,
            "dram": 2.5277921399843573,
            "package-0": 41.94205827653092,
            "psys": 58.59611608834645,
            "uncore": 5.441637351139626
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1124.9655155224127,
          "decode_tokens_per_s": 25.46414952552676,
          "end_to_end_ms": 910.2501239999998,
          "handoff_estimate_ms": 10.166610499999999,
          "handoff_plus_first_decode_ms": 48.218759,
          "output_tokens_per_s_e2e": 17.5775861800377,
          "prefill_tokens_per_s": 3138.361924210023,
          "steady_tpot_ms": 38.0521485,
          "steady_tpot_p95_ms": 41.2908715,
          "ttft_ms": 321.186665
        },
        "prefill": {
          "gpu_active_fraction": 0.78125,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 32,
          "gross_j": {
            "core": 1.862788,
            "dram": 0.633299,
            "package-0": 7.559124,
            "psys": 12.792265,
            "uncore": 4.260182
          },
          "logits": {
            "l2": 628.2510355324671,
            "sha256": "c96d07293a33b871695e60a2f64ee2de15b94f1c3e0c78181714540acc5414c9",
            "sum": -187389.82698735525,
            "top1": 5809
          },
          "net_j": {
            "core": 1.073786529098791,
            "dram": 0.4756661320357603,
            "package-0": 6.179327097577386,
            "psys": 9.602255066475285,
            "uncore": 4.2594998075452954
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.321186665,
          "watts": {
            "core": 5.7997052897572825,
            "dram": 1.9717474883336144,
            "package-0": 23.53498704561723,
            "psys": 39.828132341671164,
            "uncore": 13.263881923615978
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 518.3678398516904,
            "sha256": "f563a91e160dc6bc68c1aa9a1c95954b91970ab4b372530c63a6fe21c1c8c083",
            "sum": 92807.67067616689,
            "top1": 42
          },
          "gpu_active_fraction": 0.9811320754716981,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 53,
          "gross_j": {
            "core": 4.854113,
            "dram": 1.590694,
            "package-0": 12.479156,
            "psys": 21.629522,
            "uncore": 4.392445
          },
          "net_j": {
            "core": 3.525517169817812,
            "dram": 1.325256762173645,
            "package-0": 10.155722548142066,
            "psys": 16.257879421083487,
            "uncore": 4.391296259364995
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.5408447,
          "token_latencies_s": [
            0.043849577,
            0.038705866,
            0.038219717,
            0.037993932,
            0.037977967,
            0.037863963,
            0.038114856,
            0.03816807,
            0.037983366,
            0.038002681,
            0.039913107,
            0.037998039,
            0.038101616,
            0.037929621
          ],
          "watts": {
            "core": 8.975058829272063,
            "dram": 2.941128941450291,
            "package-0": 23.07345528208005,
            "psys": 39.99211233834778,
            "uncore": 8.121453348807893
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.1849790308624506,
    "compile_target": "AUTO:GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 401967,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.02615249902009964,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 10.018405999999999,
            "dram": 2.998649,
            "package-0": 30.825787,
            "psys": 51.754994,
            "uncore": 14.380579
          },
          "net_j": {
            "core": 5.285439166381808,
            "dram": 2.3789948263692744,
            "package-0": 23.639646245287466,
            "psys": 39.010278270589,
            "uncore": 14.378128194818315
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.936213,
            "dram": 0.131591,
            "package-0": 2.537774,
            "psys": 3.501883,
            "uncore": 0.315856
          },
          "logits": {
            "l2": 388.14070680904894,
            "sha256": "3f0cac1ef1eb2e21f28a1dc9e144e620a998951178cf91d34bb243e46b451e57",
            "sum": 39520.489759358345,
            "top1": 3862
          },
          "net_j": {
            "core": 1.7153050901533535,
            "dram": 0.10266907573207325,
            "package-0": 2.202365916699789,
            "psys": 2.907032325497897,
            "uncore": 0.3157416103857983
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.058568969,
          "watts": {
            "core": 33.0586833447589,
            "dram": 2.2467699576545392,
            "package-0": 43.32966830951045,
            "psys": 59.79075711576893,
            "uncore": 5.392889876548792
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1632.0752488184387,
          "decode_tokens_per_s": 24.502581425361424,
          "end_to_end_ms": 1254.8441019999998,
          "handoff_estimate_ms": 19.5501405,
          "handoff_plus_first_decode_ms": 58.568968999999996,
          "output_tokens_per_s_e2e": 12.750587881394052,
          "prefill_tokens_per_s": 3161.8402922538776,
          "steady_tpot_ms": 39.0188285,
          "steady_tpot_p95_ms": 42.345227900000005,
          "ttft_ms": 642.663706
        },
        "prefill": {
          "gpu_active_fraction": 0.828125,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 64,
          "gross_j": {
            "core": 3.399588,
            "dram": 1.254635,
            "package-0": 15.718648,
            "psys": 26.331964,
            "uncore": 9.458228
          },
          "logits": {
            "l2": 454.23152425564587,
            "sha256": "7b34e1c17c4e4a4835f5c0f10762b904fd00cbdbf0f7a022a5035089aee9fa90",
            "sum": -65261.15840908815,
            "top1": 8
          },
          "net_j": {
            "core": 0.9756167665852598,
            "dram": 0.9372814362950087,
            "package-0": 12.038292932864183,
            "psys": 19.80480558285045,
            "uncore": 9.456972829307293
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.642663706,
          "watts": {
            "core": 5.289839722176563,
            "dram": 1.952241877496035,
            "package-0": 24.458589855391647,
            "psys": 40.973161786111504,
            "uncore": 14.71722755104518
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 453.8179678889412,
            "sha256": "2feedb4df01c626caf4b4dd96c8d01409d1b11d9e4fa214b89bf17963116df4e",
            "sum": 5239.512074235128,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 4.682605,
            "dram": 1.612423,
            "package-0": 12.569365,
            "psys": 21.921147,
            "uncore": 4.606495
          },
          "net_j": {
            "core": 2.5945173096431953,
            "dram": 1.3390443143421926,
            "package-0": 9.398987395723493,
            "psys": 16.29844036224065,
            "uncore": 4.605413755125223
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.553611427,
          "token_latencies_s": [
            0.04218409,
            0.042644484,
            0.038901968,
            0.039041332,
            0.03937528,
            0.038623992,
            0.039055993,
            0.038714656,
            0.038832269,
            0.040446649,
            0.038846011,
            0.038996325,
            0.039128344,
            0.038795583
          ],
          "watts": {
            "core": 8.458288199314932,
            "dram": 2.9125536818083058,
            "package-0": 22.70430917243332,
            "psys": 39.59663029137584,
            "uncore": 8.320809100640187
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 9.735937,
            "dram": 3.029229,
            "package-0": 30.717878,
            "psys": 51.886525,
            "uncore": 14.47268
          },
          "net_j": {
            "core": 4.914578354441195,
            "dram": 2.3980023059800053,
            "package-0": 23.39753051523866,
            "psys": 38.90379185876414,
            "uncore": 14.470183424138245
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2100,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.442135,
            "dram": 0.124756,
            "package-0": 1.959651,
            "psys": 2.789605,
            "uncore": 0.254516
          },
          "logits": {
            "l2": 388.14070680904894,
            "sha256": "3f0cac1ef1eb2e21f28a1dc9e144e620a998951178cf91d34bb243e46b451e57",
            "sum": 39520.489759358345,
            "top1": 3862
          },
          "net_j": {
            "core": 1.2522959150342932,
            "dram": 0.09990169920885955,
            "package-0": 1.6714152203576336,
            "psys": 2.2784150303103425,
            "uncore": 0.2544176983153536
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.05033174,
          "watts": {
            "core": 28.65259575766703,
            "dram": 2.478674490490494,
            "package-0": 38.93469607845864,
            "psys": 55.42437038735398,
            "uncore": 5.056769346738261
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1602.153789107206,
          "decode_tokens_per_s": 25.44030908225242,
          "end_to_end_ms": 1278.279285,
          "handoff_estimate_ms": 12.049381499999997,
          "handoff_plus_first_decode_ms": 50.331739999999996,
          "output_tokens_per_s_e2e": 12.516826477400047,
          "prefill_tokens_per_s": 2950.6414876831186,
          "steady_tpot_ms": 38.2823585,
          "steady_tpot_p95_ms": 39.660728899999995,
          "ttft_ms": 688.6638069999999
        },
        "prefill": {
          "gpu_active_fraction": 0.7647058823529411,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 3.684988,
            "dram": 1.30322,
            "package-0": 16.338703,
            "psys": 27.476248,
            "uncore": 9.666845
          },
          "logits": {
            "l2": 454.23152425564587,
            "sha256": "7b34e1c17c4e4a4835f5c0f10762b904fd00cbdbf0f7a022a5035089aee9fa90",
            "sum": -65261.15840908815,
            "top1": 8
          },
          "net_j": {
            "core": 1.0875155706764739,
            "dram": 0.9631511408039959,
            "package-0": 12.394918236705006,
            "psys": 20.481893530500358,
            "uncore": 9.665499987640342
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.688663807,
          "watts": {
            "core": 5.350924446070099,
            "dram": 1.8923892714460602,
            "package-0": 23.725223881266057,
            "psys": 39.897912044621215,
            "uncore": 14.03710330315065
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 453.8179678889412,
            "sha256": "2feedb4df01c626caf4b4dd96c8d01409d1b11d9e4fa214b89bf17963116df4e",
            "sum": 5239.512074235128,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 53,
          "gross_j": {
            "core": 4.608814,
            "dram": 1.601253,
            "package-0": 12.419524,
            "psys": 21.620672,
            "uncore": 4.551319
          },
          "net_j": {
            "core": 2.574766868730428,
            "dram": 1.3349494659671497,
            "package-0": 9.33119705817602,
            "psys": 16.143483297953438,
            "uncore": 4.55026573818255
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.539283738,
          "token_latencies_s": [
            0.041746276,
            0.038537742,
            0.038375251,
            0.038293233,
            0.0380478,
            0.038220112,
            0.038149965,
            0.038396353,
            0.038101786,
            0.03805424,
            0.038533271,
            0.038417281,
            0.038271484,
            0.038116643
          ],
          "watts": {
            "core": 8.546176484186883,
            "dram": 2.9692217420433327,
            "package-0": 23.029665322487435,
            "psys": 40.091459238476055,
            "uncore": 8.43956284845363
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 9.396459,
            "dram": 3.032829,
            "package-0": 30.241073,
            "psys": 51.64977,
            "uncore": 14.374719
          },
          "net_j": {
            "core": 4.544882830167648,
            "dram": 2.397646137324883,
            "package-0": 22.874845755180708,
            "psys": 38.58566849713154,
            "uncore": 14.372206777025626
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.625606,
            "dram": 0.126465,
            "package-0": 2.132929,
            "psys": 3.011039,
            "uncore": 0.239502
          },
          "logits": {
            "l2": 388.14070680904894,
            "sha256": "3f0cac1ef1eb2e21f28a1dc9e144e620a998951178cf91d34bb243e46b451e57",
            "sum": 39520.489759358345,
            "top1": 3862
          },
          "net_j": {
            "core": 1.4276382132655119,
            "dram": 0.10054646526125459,
            "package-0": 1.8323512798545583,
            "psys": 2.4779604354956857,
            "uncore": 0.23939948914484477
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.05248689,
          "watts": {
            "core": 30.97165787494744,
            "dram": 2.4094588191451236,
            "package-0": 40.63736677863748,
            "psys": 57.367449281144296,
            "uncore": 4.5630823239860465
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1592.1749452598515,
          "decode_tokens_per_s": 24.565094490356536,
          "end_to_end_ms": 1286.29081,
          "handoff_estimate_ms": 12.980355500000005,
          "handoff_plus_first_decode_ms": 52.48689,
          "output_tokens_per_s_e2e": 12.43886675984259,
          "prefill_tokens_per_s": 3007.3928910139794,
          "steady_tpot_ms": 39.506534499999994,
          "steady_tpot_p95_ms": 42.16395955,
          "ttft_ms": 675.668286
        },
        "prefill": {
          "gpu_active_fraction": 0.7910447761194029,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 67,
          "gross_j": {
            "core": 3.690603,
            "dram": 1.288509,
            "package-0": 16.16335,
            "psys": 27.319694,
            "uncore": 9.536963
          },
          "logits": {
            "l2": 454.23152425564587,
            "sha256": "7b34e1c17c4e4a4835f5c0f10762b904fd00cbdbf0f7a022a5035089aee9fa90",
            "sum": -65261.15840908815,
            "top1": 8
          },
          "net_j": {
            "core": 1.142146515988371,
            "dram": 0.9548574549654291,
            "package-0": 12.293986946093433,
            "psys": 20.457327416992882,
            "uncore": 9.535643368873082
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.675668286,
          "watts": {
            "core": 5.462152178028969,
            "dram": 1.907014176480084,
            "package-0": 23.922019628430515,
            "psys": 40.43358933084511,
            "uncore": 14.114859610267398
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 453.8179678889412,
            "sha256": "2feedb4df01c626caf4b4dd96c8d01409d1b11d9e4fa214b89bf17963116df4e",
            "sum": 5239.512074235128,
            "top1": 42
          },
          "gpu_active_fraction": 0.9818181818181818,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 4.08025,
            "dram": 1.617855,
            "package-0": 11.944794,
            "psys": 21.319037,
            "uncore": 4.598254
          },
          "net_j": {
            "core": 1.9750981009137654,
            "dram": 1.3422422170981994,
            "package-0": 8.748507529232718,
            "psys": 15.650380644642972,
            "uncore": 4.5971639190077
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.558135634,
          "token_latencies_s": [
            0.044164889,
            0.040065096,
            0.038947973,
            0.040946505,
            0.040507038,
            0.038612673,
            0.03844799,
            0.04072849,
            0.041086536,
            0.040085377,
            0.038408862,
            0.038505675,
            0.038718139,
            0.038889417
          ],
          "watts": {
            "core": 7.3104990103534595,
            "dram": 2.898677134096047,
            "package-0": 21.401238825041585,
            "psys": 38.19687491947522,
            "uncore": 8.238595996900639
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.8732490157708526,
    "compile_target": "NPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402038,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 2048,
      "NPUW_LLM_MIN_RESPONSE_LEN": 2048,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.02616929356008768,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 14.178613,
            "dram": 3.661306,
            "package-0": 34.775546,
            "psys": 61.267604,
            "uncore": 10.533786000000001
          },
          "net_j": {
            "core": 7.855560562014698,
            "dram": 2.8334730632776717,
            "package-0": 25.175152332323574,
            "psys": 44.241178779378494,
            "uncore": 10.531650949636026
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 19,
          "gross_j": {
            "core": 3.304191,
            "dram": 0.23059,
            "package-0": 3.934072,
            "psys": 6.120712,
            "uncore": 0.005432
          },
          "logits": {
            "l2": 1026.4886340623634,
            "sha256": "2b5d51c7401796ac587631de9e5f269ea1f06106fe3f8dad834e85e69bf2f811",
            "sum": -244601.09724290296,
            "top1": 370
          },
          "net_j": {
            "core": 2.5832979539685184,
            "dram": 0.13620852630323968,
            "package-0": 2.839528327925412,
            "psys": 4.179524312454586,
            "uncore": 0.005058710164097304
          },
          "npu_busy_delta_us": 56471,
          "npu_utilization": 0.2954597524260638,
          "seconds": 0.191129247,
          "watts": {
            "core": 17.287730956215192,
            "dram": 1.2064610917448966,
            "package-0": 20.583307169101126,
            "psys": 32.02394241630638,
            "uncore": 0.028420558785542643
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1221.6501600116544,
          "decode_tokens_per_s": 19.37041679877237,
          "end_to_end_ms": 1676.421014,
          "handoff_estimate_ms": 151.2270395,
          "handoff_plus_first_decode_ms": 191.129247,
          "output_tokens_per_s_e2e": 9.54414187509105,
          "prefill_tokens_per_s": 2252.6610209608502,
          "steady_tpot_ms": 39.902207499999996,
          "steady_tpot_p95_ms": 48.3112845,
          "ttft_ms": 902.044285
        },
        "prefill": {
          "gpu_active_fraction": 0.9888888888888889,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 90,
          "gross_j": {
            "core": 6.016403,
            "dram": 1.742061,
            "package-0": 20.765633,
            "psys": 36.900967,
            "uncore": 10.528354
          },
          "logits": {
            "l2": 366.74491757063316,
            "sha256": "1f0df7529225c1c78b67b5a87bd8d4ea0b140bf8c9c4e37ddaa073fb1bd6884e",
            "sum": -50010.3206056105,
            "top1": 90
          },
          "net_j": {
            "core": 2.6141112917668807,
            "dram": 1.2966227936848753,
            "package-0": 15.599877996121418,
            "psys": 27.739432137349148,
            "uncore": 10.52659223947193
          },
          "npu_busy_delta_us": 9726,
          "npu_utilization": 0.010782175733201392,
          "seconds": 0.902044285,
          "watts": {
            "core": 6.669742384100355,
            "dram": 1.9312366687185432,
            "package-0": 23.020635843837756,
            "psys": 40.90815452591665,
            "uncore": 11.671659778876599
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 813.1332002581323,
            "sha256": "5e3c7b11b2975536d5ab3fd0541e09bbe1892c78109e321d44e4f8d6ce291c4f",
            "sum": -91797.23669979721,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 4.858019,
            "dram": 1.688655,
            "package-0": 10.075841,
            "psys": 18.245925,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.658151316279299,
            "dram": 1.4006417432895568,
            "package-0": 6.735746008276745,
            "psys": 12.32222232957476,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 571127,
          "npu_utilization": 0.9792189724361298,
          "seconds": 0.583247482,
          "token_latencies_s": [
            0.053266943,
            0.045642853,
            0.043085412,
            0.041374964,
            0.040419602,
            0.039826308,
            0.039855048,
            0.039860939,
            0.039777561,
            0.039593654,
            0.039838664,
            0.039907398,
            0.039897017,
            0.040861054
          ],
          "watts": {
            "core": 8.329258419327388,
            "dram": 2.895263249503407,
            "package-0": 17.275412772377816,
            "psys": 31.283332655690746,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 12.49124,
            "dram": 3.5721950000000002,
            "package-0": 32.787330000000004,
            "psys": 57.997777,
            "uncore": 10.48325
          },
          "net_j": {
            "core": 6.587447668510164,
            "dram": 2.7992528503800553,
            "package-0": 23.823505791613226,
            "psys": 42.1003161397483,
            "uncore": 10.481335872016974
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.125,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.360773,
            "dram": 0.15509,
            "package-0": 2.765984,
            "psys": 3.938162,
            "uncore": 0.007996
          },
          "logits": {
            "l2": 957.03468159783,
            "sha256": "a85018f448b20deb48b5ea1269f1000c3280c7556b9fea050e8873cb46a40981",
            "sum": -197233.62805647403,
            "top1": 83
          },
          "net_j": {
            "core": 2.050858500705465,
            "dram": 0.11451506683282053,
            "package-0": 2.2954357599325212,
            "psys": 3.103638469034935,
            "uncore": 0.00783552136419905
          },
          "npu_busy_delta_us": 54586,
          "npu_utilization": 0.6643287736399075,
          "seconds": 0.082167147,
          "watts": {
            "core": 28.73134928245714,
            "dram": 1.8874940370024045,
            "package-0": 33.662894489935255,
            "psys": 47.92866910664429,
            "uncore": 0.09731383274144835
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1308.4061208293747,
          "decode_tokens_per_s": 22.47622076528829,
          "end_to_end_ms": 1565.263237,
          "handoff_estimate_ms": 42.030404499999996,
          "handoff_plus_first_decode_ms": 82.167147,
          "output_tokens_per_s_e2e": 10.22192281897949,
          "prefill_tokens_per_s": 2263.0802880466854,
          "steady_tpot_ms": 40.136742500000004,
          "steady_tpot_p95_ms": 48.466140249999995,
          "ttft_ms": 897.891255
        },
        "prefill": {
          "gpu_active_fraction": 0.9887640449438202,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 89,
          "gross_j": {
            "core": 6.207138,
            "dram": 1.73437,
            "package-0": 20.904243,
            "psys": 37.129239,
            "uncore": 10.475254
          },
          "logits": {
            "l2": 369.2233804425175,
            "sha256": "ea02272bdbf413a702ab91b2f85a27d9af2a58670dcde9962d3526c27e15f262",
            "sum": -46712.414185937494,
            "top1": 90
          },
          "net_j": {
            "core": 2.8205105111051174,
            "dram": 1.2909826001305618,
            "package-0": 15.762271235880824,
            "psys": 28.009884027808333,
            "uncore": 10.473500350652774
          },
          "npu_busy_delta_us": 7980,
          "npu_utilization": 0.008887490501285704,
          "seconds": 0.897891255,
          "watts": {
            "core": 6.913017545760594,
            "dram": 1.9316036216434695,
            "package-0": 23.28148635326669,
            "psys": 41.35159886371763,
            "uncore": 11.666506318741238
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 871.4172103989001,
            "sha256": "160a3b8ab6ee81bb0777c0eee19112601ae899ac4cee1eb1b9d3e5204371b1fc",
            "sum": -145708.02170770615,
            "top1": 75
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 3.923329,
            "dram": 1.682735,
            "package-0": 9.117103,
            "psys": 16.930376,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.716078656699581,
            "dram": 1.3937551834166728,
            "package-0": 5.7657987957998795,
            "psys": 10.986793642905036,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 572066,
          "npu_utilization": 0.9775483143436434,
          "seconds": 0.585204835,
          "token_latencies_s": [
            0.052923031,
            0.046066276,
            0.043457531,
            0.041763888,
            0.040699016,
            0.040042184,
            0.040006913,
            0.040116214,
            0.04002152,
            0.040157271,
            0.04023111,
            0.03985848,
            0.039904702,
            0.039931882
          ],
          "watts": {
            "core": 6.704197855781557,
            "dram": 2.8754632555282975,
            "package-0": 15.579336421579635,
            "psys": 28.93068373230375,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.188255999999999,
            "dram": 3.587881,
            "package-0": 34.45908,
            "psys": 59.431549,
            "uncore": 10.424778
          },
          "net_j": {
            "core": 8.204117462763234,
            "dram": 2.804419684817169,
            "package-0": 25.373264836700248,
            "psys": 43.31773556678963,
            "uncore": 10.42301598782402
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.4096,
            "dram": 0.163329,
            "package-0": 2.839531,
            "psys": 4.119862,
            "uncore": 0.0
          },
          "logits": {
            "l2": 837.2293686109759,
            "sha256": "894545dfa5886dde4385e1e4378f9e7553287f9507144db59a71f97471aef493",
            "sum": -125513.81106831506,
            "top1": 62
          },
          "net_j": {
            "core": 2.0252021703778436,
            "dram": 0.11300248645503379,
            "package-0": 2.2558935044126684,
            "psys": 3.084773177754483,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 54364,
          "npu_utilization": 0.5334259806593735,
          "seconds": 0.101914796,
          "watts": {
            "core": 23.64327943118289,
            "dram": 1.6026034139341259,
            "package-0": 27.861813116909932,
            "psys": 40.424571913974106,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1290.838768948975,
          "decode_tokens_per_s": 21.91725837821502,
          "end_to_end_ms": 1586.565301,
          "handoff_estimate_ms": 62.070555999999996,
          "handoff_plus_first_decode_ms": 101.914796,
          "output_tokens_per_s_e2e": 10.084677882413867,
          "prefill_tokens_per_s": 2252.339299326418,
          "steady_tpot_ms": 39.844240000000006,
          "steady_tpot_p95_ms": 48.58107759999999,
          "ttft_ms": 902.173132
        },
        "prefill": {
          "gpu_active_fraction": 0.9888888888888889,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 90,
          "gross_j": {
            "core": 6.945905,
            "dram": 1.736751,
            "package-0": 21.591986,
            "psys": 37.487758,
            "uncore": 10.424778
          },
          "logits": {
            "l2": 369.6101676549024,
            "sha256": "e185f24df452e75d08c9dbbd0c65b619de2470e823b0c7928125bccf1bc0ffdb",
            "sum": -39722.19562444929,
            "top1": 90
          },
          "net_j": {
            "core": 3.5431273122124165,
            "dram": 1.2912491677880942,
            "package-0": 16.425493125435846,
            "psys": 28.32491451392336,
            "uncore": 10.42301598782402
          },
          "npu_busy_delta_us": 9348,
          "npu_utilization": 0.010361647524657163,
          "seconds": 0.902173132,
          "watts": {
            "core": 7.699082087051114,
            "dram": 1.925075064195106,
            "package-0": 23.933306406646565,
            "psys": 41.55273158810941,
            "uncore": 11.555185618185755
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 906.6597247212674,
            "sha256": "2f8dbeac505661241a401ff006629454b54aeca37d59486d1e910a10ea22d139",
            "sum": -173918.8459377624,
            "top1": 62
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 57,
          "gross_j": {
            "core": 4.832751,
            "dram": 1.687801,
            "package-0": 10.027563,
            "psys": 17.823929,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.6357879801729744,
            "dram": 1.4001680305740412,
            "package-0": 6.691878206851732,
            "psys": 11.908047875111784,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 571930,
          "npu_utilization": 0.9818922184982455,
          "seconds": 0.582477373,
          "token_latencies_s": [
            0.053674813,
            0.045838297,
            0.042991903,
            0.041648133,
            0.040481749,
            0.039860386,
            0.039828094,
            0.039559582,
            0.039788167,
            0.03990437,
            0.039754667,
            0.039558386,
            0.039766189,
            0.039798849
          ],
          "watts": {
            "core": 8.296890530029223,
            "dram": 2.89762500353812,
            "package-0": 17.21536915391905,
            "psys": 30.600208396421262,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.2246395656839013,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 401877,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.02877042442560196,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.224703,
            "dram": 2.95507,
            "package-0": 31.846659,
            "psys": 53.059556,
            "uncore": 14.227747
          },
          "net_j": {
            "core": 6.496222690010083,
            "dram": 2.3360032154613775,
            "package-0": 24.667330207303156,
            "psys": 40.32692137537289,
            "uncore": 14.225298518011304
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.061884,
            "dram": 0.133605,
            "package-0": 2.627556,
            "psys": 3.645803,
            "uncore": 0.276366
          },
          "logits": {
            "l2": 388.14070680904894,
            "sha256": "3f0cac1ef1eb2e21f28a1dc9e144e620a998951178cf91d34bb243e46b451e57",
            "sum": 39520.489759358345,
            "top1": 3862
          },
          "net_j": {
            "core": 1.8320066260658108,
            "dram": 0.10350876666707012,
            "package-0": 2.2785294330523658,
            "psys": 3.0267997645230493,
            "uncore": 0.2762469658552458
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.060947029,
          "watts": {
            "core": 33.830754900292185,
            "dram": 2.192149513965644,
            "package-0": 43.1121261054415,
            "psys": 59.819207922341874,
            "uncore": 4.534527843842889
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1633.6238106579526,
          "decode_tokens_per_s": 24.215759902290376,
          "end_to_end_ms": 1253.6545970000002,
          "handoff_estimate_ms": 21.786800999999993,
          "handoff_plus_first_decode_ms": 60.947029,
          "output_tokens_per_s_e2e": 12.762686020765255,
          "prefill_tokens_per_s": 3203.9189630380047,
          "steady_tpot_ms": 39.160228000000004,
          "steady_tpot_p95_ms": 42.22771325,
          "ttft_ms": 634.223282
        },
        "prefill": {
          "gpu_active_fraction": 0.8253968253968254,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 3.766897,
            "dram": 1.209347,
            "package-0": 15.950277,
            "psys": 26.54284,
            "uncore": 9.389808
          },
          "logits": {
            "l2": 454.23152425564587,
            "sha256": "7b34e1c17c4e4a4835f5c0f10762b904fd00cbdbf0f7a022a5035089aee9fa90",
            "sum": -65261.15840908815,
            "top1": 8
          },
          "net_j": {
            "core": 1.374760993598723,
            "dram": 0.8961613994312079,
            "package-0": 12.318257877382536,
            "psys": 20.101406020128554,
            "uncore": 9.388569314092011
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.634223282,
          "watts": {
            "core": 5.9393861860782335,
            "dram": 1.9068158396619692,
            "package-0": 25.14930853642172,
            "psys": 41.85093917760023,
            "uncore": 14.805208617365139
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 453.8179678889412,
            "sha256": "2feedb4df01c626caf4b4dd96c8d01409d1b11d9e4fa214b89bf17963116df4e",
            "sum": 5239.512074235128,
            "top1": 42
          },
          "gpu_active_fraction": 0.9821428571428571,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 56,
          "gross_j": {
            "core": 5.395922,
            "dram": 1.612118,
            "package-0": 13.268826,
            "psys": 22.870913,
            "uncore": 4.561573
          },
          "net_j": {
            "core": 3.289455070345549,
            "dram": 1.3363330493630994,
            "package-0": 10.070542896868254,
            "psys": 17.198715590721285,
            "uncore": 4.560482238064046
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.558484286,
          "token_latencies_s": [
            0.044595803,
            0.039527306,
            0.039287647,
            0.038826702,
            0.040952588,
            0.040452623,
            0.038780344,
            0.038644735,
            0.038873274,
            0.039032809,
            0.038856582,
            0.038941791,
            0.040882787,
            0.040789719
          ],
          "watts": {
            "core": 9.661725737436415,
            "dram": 2.8865950939217653,
            "package-0": 23.75863803623653,
            "psys": 40.95175741435275,
            "uncore": 8.167773228985713
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 11.064912,
            "dram": 3.010735,
            "package-0": 31.979288,
            "psys": 53.449387,
            "uncore": 14.415795999999999
          },
          "net_j": {
            "core": 6.288190086378872,
            "dram": 2.385352280428749,
            "package-0": 24.726713193520734,
            "psys": 40.58684960059669,
            "uncore": 14.413322537744948
          }
        },
        "first_decode": {
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.690059,
            "dram": 0.124084,
            "package-0": 2.213617,
            "psys": 3.085869,
            "uncore": 0.255371
          },
          "logits": {
            "l2": 388.14070680904894,
            "sha256": "3f0cac1ef1eb2e21f28a1dc9e144e620a998951178cf91d34bb243e46b451e57",
            "sum": 39520.489759358345,
            "top1": 3862
          },
          "net_j": {
            "core": 1.499058843808815,
            "dram": 0.09907768829138602,
            "package-0": 1.923618346974861,
            "psys": 2.571552551036241,
            "uncore": 0.2552720970942854
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.050639573,
          "watts": {
            "core": 33.37427430519605,
            "dram": 2.450336617174872,
            "package-0": 43.71318454837682,
            "psys": 60.937895349157074,
            "uncore": 5.042913770224722
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1617.1253345520813,
          "decode_tokens_per_s": 25.248461695390443,
          "end_to_end_ms": 1266.4448180000002,
          "handoff_estimate_ms": 12.310220000000003,
          "handoff_plus_first_decode_ms": 50.639573,
          "output_tokens_per_s_e2e": 12.633791676188135,
          "prefill_tokens_per_s": 3022.238957969607,
          "steady_tpot_ms": 38.329353,
          "steady_tpot_p95_ms": 41.38146835,
          "ttft_ms": 672.3492180000001
        },
        "prefill": {
          "gpu_active_fraction": 0.7910447761194029,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 67,
          "gross_j": {
            "core": 4.691333,
            "dram": 1.283139,
            "package-0": 17.299028,
            "psys": 28.634265,
            "uncore": 9.649389
          },
          "logits": {
            "l2": 454.23152425564587,
            "sha256": "7b34e1c17c4e4a4835f5c0f10762b904fd00cbdbf0f7a022a5035089aee9fa90",
            "sum": -65261.15840908815,
            "top1": 8
          },
          "net_j": {
            "core": 2.1553952334429822,
            "dram": 0.9511264428966679,
            "package-0": 13.448672319028802,
            "psys": 21.80560824443612,
            "uncore": 9.648075851263236
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.672349218,
          "watts": {
            "core": 6.9775242900632035,
            "dram": 1.9084412767176,
            "package-0": 25.729230490456224,
            "psys": 42.58838150385117,
            "uncore": 14.351751651773318
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 453.8179678889412,
            "sha256": "2feedb4df01c626caf4b4dd96c8d01409d1b11d9e4fa214b89bf17963116df4e",
            "sum": 5239.512074235128,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 54,
          "gross_j": {
            "core": 4.68352,
            "dram": 1.603512,
            "package-0": 12.466643,
            "psys": 21.729253,
            "uncore": 4.511036
          },
          "net_j": {
            "core": 2.6337360091270745,
            "dram": 1.335148149240695,
            "package-0": 9.354422527517071,
            "psys": 16.20968880512433,
            "uncore": 4.509974589387426
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.543456027,
          "token_latencies_s": [
            0.043760027,
            0.040100706,
            0.038354925,
            0.038688037,
            0.038334887,
            0.038244146,
            0.038476487,
            0.038410052,
            0.038095727,
            0.038323819,
            0.038280581,
            0.037913214,
            0.03821631,
            0.038232035
          ],
          "watts": {
            "core": 8.618029366339146,
            "dram": 2.9505827892860967,
            "package-0": 22.939561584805094,
            "psys": 39.9834612561947,
            "uncore": 8.300645822076788
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 12.121491,
            "dram": 3.017693,
            "package-0": 33.008766,
            "psys": 54.76048,
            "uncore": 14.337121
          },
          "net_j": {
            "core": 7.316822382260864,
            "dram": 2.388651414340159,
            "package-0": 25.713759256749945,
            "psys": 41.82268899919216,
            "uncore": 14.334633066498967
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.616695,
            "dram": 0.124817,
            "package-0": 2.132258,
            "psys": 2.997795,
            "uncore": 0.251159
          },
          "logits": {
            "l2": 388.14070680904894,
            "sha256": "3f0cac1ef1eb2e21f28a1dc9e144e620a998951178cf91d34bb243e46b451e57",
            "sum": 39520.489759358345,
            "top1": 3862
          },
          "net_j": {
            "core": 1.4246279458082924,
            "dram": 0.09967100681418523,
            "package-0": 1.8406394583358323,
            "psys": 2.480605657097421,
            "uncore": 0.25105954463760444
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.050922438,
          "watts": {
            "core": 31.748185348077797,
            "dram": 2.451119877646078,
            "package-0": 41.872661320732526,
            "psys": 58.86982473227224,
            "uncore": 4.932187260947718
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1607.7192075447776,
          "decode_tokens_per_s": 25.110921889340272,
          "end_to_end_ms": 1273.854284,
          "handoff_estimate_ms": 12.282993499999998,
          "handoff_plus_first_decode_ms": 50.922438,
          "output_tokens_per_s_e2e": 12.560306308943575,
          "prefill_tokens_per_s": 3003.6748453983278,
          "steady_tpot_ms": 38.6394445,
          "steady_tpot_p95_ms": 40.783110050000005,
          "ttft_ms": 676.5046500000001
        },
        "prefill": {
          "gpu_active_fraction": 0.7910447761194029,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 67,
          "gross_j": {
            "core": 5.221911,
            "dram": 1.284237,
            "package-0": 17.708512,
            "psys": 29.153307,
            "uncore": 9.505408
          },
          "logits": {
            "l2": 454.23152425564587,
            "sha256": "7b34e1c17c4e4a4835f5c0f10762b904fd00cbdbf0f7a022a5035089aee9fa90",
            "sum": -65261.15840908815,
            "top1": 8
          },
          "net_j": {
            "core": 2.6702999543447987,
            "dram": 0.9501724503201123,
            "package-0": 13.83435932368807,
            "psys": 22.282445958271417,
            "uncore": 9.504086735391104
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.67650465,
          "watts": {
            "core": 7.718958029335053,
            "dram": 1.898341718715459,
            "package-0": 26.17648230503663,
            "psys": 43.09402307877706,
            "uncore": 14.050765209078753
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 453.8179678889412,
            "sha256": "2feedb4df01c626caf4b4dd96c8d01409d1b11d9e4fa214b89bf17963116df4e",
            "sum": 5239.512074235128,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 54,
          "gross_j": {
            "core": 5.282885,
            "dram": 1.608639,
            "package-0": 13.167996,
            "psys": 22.609378,
            "uncore": 4.580554
          },
          "net_j": {
            "core": 3.221894482107773,
            "dram": 1.3388079572058613,
            "package-0": 10.038760474726045,
            "psys": 17.059637383823322,
            "uncore": 4.579486786470259
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.546427196,
          "token_latencies_s": [
            0.042838841,
            0.039676178,
            0.039318594,
            0.038500786,
            0.038632513,
            0.038396205,
            0.038521508,
            0.038862444,
            0.038765446,
            0.038699566,
            0.038519141,
            0.038460859,
            0.038561309,
            0.038646376
          ],
          "watts": {
            "core": 9.668049172281682,
            "dram": 2.943921920020979,
            "package-0": 24.098353991882938,
            "psys": 41.37674362752618,
            "uncore": 8.382734302997614
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.166511282324791,
    "compile_target": "AUTO:GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402375,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.026095177978277206,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 18.209975,
            "dram": 4.7343020000000005,
            "package-0": 57.792576000000004,
            "psys": 95.677123,
            "uncore": 29.015367
          },
          "net_j": {
            "core": 9.586043665621498,
            "dram": 3.6052311164909274,
            "package-0": 44.69872036855377,
            "psys": 72.45499675437996,
            "uncore": 29.010901391779242
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.146906,
            "dram": 0.139648,
            "package-0": 2.79876,
            "psys": 3.904348,
            "uncore": 0.342651
          },
          "logits": {
            "l2": 386.3595092072282,
            "sha256": "a08f57e55d540bd7c46befedb70a9eeed4a0b576f2871d968449003d95edde09",
            "sum": 34339.569468914415,
            "top1": 3862
          },
          "net_j": {
            "core": 1.8948035148991353,
            "dram": 0.106641988640312,
            "package-0": 2.4159886841987666,
            "psys": 3.2254980058553375,
            "uncore": 0.34252045734070813
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.066839538,
          "watts": {
            "core": 32.120299814160894,
            "dram": 2.089302292903341,
            "package-0": 41.872820844452875,
            "psys": 58.413749059725696,
            "uncore": 5.126471700028806
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1791.4238235718765,
          "decode_tokens_per_s": 23.23151177517207,
          "end_to_end_ms": 2286.449441,
          "handoff_estimate_ms": 26.066698499999998,
          "handoff_plus_first_decode_ms": 66.839538,
          "output_tokens_per_s_e2e": 6.9977493108276425,
          "prefill_tokens_per_s": 2486.630225771123,
          "steady_tpot_ms": 40.7728395,
          "steady_tpot_p95_ms": 43.9999003,
          "ttft_ms": 1640.774715
        },
        "prefill": {
          "gpu_active_fraction": 0.8588957055214724,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 163,
          "gross_j": {
            "core": 10.768588,
            "dram": 2.932793,
            "package-0": 41.390213,
            "psys": 68.230477,
            "uncore": 23.778137
          },
          "logits": {
            "l2": 388.37710963475644,
            "sha256": "cc21758ed0d374d7022ec1be971528d81a755ceb6dd836aac0bc64de1d25bf88",
            "sum": -1062.9958727225894,
            "top1": 72
          },
          "net_j": {
            "core": 4.579984734337649,
            "dram": 2.122562551040601,
            "package-0": 31.993955703997393,
            "psys": 51.56609033606972,
            "uncore": 23.77493244322112
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.640774715,
          "watts": {
            "core": 6.563111865116716,
            "dram": 1.7874440489534238,
            "package-0": 25.22601830806492,
            "psys": 41.584305496809165,
            "uncore": 14.492018180570268
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 450.3147795636656,
            "sha256": "4f4fcbcd85ba65dc12597e879daaf0bd341940cd416489da4580479054b6afca",
            "sum": 3555.901248254231,
            "top1": 42
          },
          "gpu_active_fraction": 0.9827586206896551,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 5.294481,
            "dram": 1.661861,
            "package-0": 13.603603,
            "psys": 23.542298,
            "uncore": 4.894579
          },
          "net_j": {
            "core": 3.1112554163847137,
            "dram": 1.3760265768100144,
            "package-0": 10.28877598035761,
            "psys": 17.663408412454903,
            "uncore": 4.893448491217411
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.578835188,
          "token_latencies_s": [
            0.046730757,
            0.041800779,
            0.041028553,
            0.041001933,
            0.041178261,
            0.042529439,
            0.040810324,
            0.040505382,
            0.040519514,
            0.04018004,
            0.040700433,
            0.040497106,
            0.040735355,
            0.040589902
          ],
          "watts": {
            "core": 9.146784973963953,
            "dram": 2.8710434929536452,
            "package-0": 23.50168628656349,
            "psys": 40.67185010182898,
            "uncore": 8.455911287825854
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 17.817156,
            "dram": 4.819262,
            "package-0": 57.560399,
            "psys": 97.004818,
            "uncore": 28.963671
          },
          "net_j": {
            "core": 8.98013209621303,
            "dram": 3.662292399967974,
            "package-0": 44.14300144413635,
            "psys": 73.20888586969251,
            "uncore": 28.95909504907209
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.945186,
            "dram": 0.134643,
            "package-0": 2.493219,
            "psys": 3.485038,
            "uncore": 0.257018
          },
          "logits": {
            "l2": 386.3595092072282,
            "sha256": "a08f57e55d540bd7c46befedb70a9eeed4a0b576f2871d968449003d95edde09",
            "sum": 34339.569468914415,
            "top1": 3862
          },
          "net_j": {
            "core": 1.719660818573979,
            "dram": 0.10511656872715963,
            "package-0": 2.150800437936988,
            "psys": 2.877754148635842,
            "uncore": 0.2569012194894515
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.059793139,
          "watts": {
            "core": 32.53192644728018,
            "dram": 2.2518135400116726,
            "package-0": 41.697409463651,
            "psys": 58.28491459530164,
            "uncore": 4.298453038232363
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1748.226123800929,
          "decode_tokens_per_s": 23.853103467420855,
          "end_to_end_ms": 2342.9463410000003,
          "handoff_estimate_ms": 19.824571000000006,
          "handoff_plus_first_decode_ms": 59.793139000000004,
          "output_tokens_per_s_e2e": 6.829008296097379,
          "prefill_tokens_per_s": 2380.2615383773136,
          "steady_tpot_ms": 39.968568,
          "steady_tpot_p95_ms": 43.72544909999999,
          "ttft_ms": 1714.097352
        },
        "prefill": {
          "gpu_active_fraction": 0.8352941176470589,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 170,
          "gross_j": {
            "core": 10.838352,
            "dram": 3.028801,
            "package-0": 41.799819,
            "psys": 70.456912,
            "uncore": 23.866699
          },
          "logits": {
            "l2": 388.37710963475644,
            "sha256": "cc21758ed0d374d7022ec1be971528d81a755ceb6dd836aac0bc64de1d25bf88",
            "sum": -1062.9958727225894,
            "top1": 72
          },
          "net_j": {
            "core": 4.373193576193849,
            "dram": 2.182363122590927,
            "package-0": 31.98366349580651,
            "psys": 53.0478303069371,
            "uncore": 23.863351238580478
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.714097352,
          "watts": {
            "core": 6.323066765930107,
            "dram": 1.7669947371810653,
            "package-0": 24.385907224714035,
            "psys": 41.10438180059682,
            "uncore": 13.92377099944321
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 450.3147795636656,
            "sha256": "4f4fcbcd85ba65dc12597e879daaf0bd341940cd416489da4580479054b6afca",
            "sum": 3555.901248254231,
            "top1": 42
          },
          "gpu_active_fraction": 0.9824561403508771,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 57,
          "gross_j": {
            "core": 5.033618,
            "dram": 1.655818,
            "package-0": 13.267361,
            "psys": 23.062868,
            "uncore": 4.839954
          },
          "net_j": {
            "core": 2.887277701445201,
            "dram": 1.3748127086498876,
            "package-0": 10.008537510392856,
            "psys": 17.28330141411956,
            "uncore": 4.838842591002161
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.56905585,
          "token_latencies_s": [
            0.045728922,
            0.042646656,
            0.040019732,
            0.039950196,
            0.04007741,
            0.039750512,
            0.039966918,
            0.039837886,
            0.039718784,
            0.039806412,
            0.039970218,
            0.039930786,
            0.040127735,
            0.041494791
          ],
          "watts": {
            "core": 8.845560589527372,
            "dram": 2.9097636022896523,
            "package-0": 23.314690464916577,
            "psys": 40.52830315337238,
            "uncore": 8.505235470296983
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 17.309892,
            "dram": 4.800769,
            "package-0": 56.924414999999996,
            "psys": 95.3703,
            "uncore": 28.884141999999997
          },
          "net_j": {
            "core": 8.487423931997183,
            "dram": 3.6457050935237088,
            "package-0": 43.52911780706399,
            "psys": 71.61356315581945,
            "uncore": 28.879573586314375
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.722103,
            "dram": 0.13507,
            "package-0": 2.239802,
            "psys": 3.226737,
            "uncore": 0.258605
          },
          "logits": {
            "l2": 386.3595092072282,
            "sha256": "a08f57e55d540bd7c46befedb70a9eeed4a0b576f2871d968449003d95edde09",
            "sum": 34339.569468914415,
            "top1": 3862
          },
          "net_j": {
            "core": 1.4944962651770213,
            "dram": 0.10527104553073696,
            "package-0": 1.8942229814569767,
            "psys": 2.6138480372432618,
            "uncore": 0.25848714162813724
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.060345018,
          "watts": {
            "core": 28.537616808731418,
            "dram": 2.238295794360356,
            "package-0": 37.11660173835726,
            "psys": 53.47147298887209,
            "uncore": 4.285440763312059
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1751.1104518795883,
          "decode_tokens_per_s": 23.812829937474554,
          "end_to_end_ms": 2339.0871750000006,
          "handoff_estimate_ms": 20.376389500000002,
          "handoff_plus_first_decode_ms": 60.345018,
          "output_tokens_per_s_e2e": 6.840275202654642,
          "prefill_tokens_per_s": 2387.117089941522,
          "steady_tpot_ms": 39.9686285,
          "steady_tpot_p95_ms": 43.48767765,
          "ttft_ms": 1709.174643
        },
        "prefill": {
          "gpu_active_fraction": 0.834319526627219,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 169,
          "gross_j": {
            "core": 10.529025,
            "dram": 3.011589,
            "package-0": 41.391861,
            "psys": 69.304388,
            "uncore": 23.789246
          },
          "logits": {
            "l2": 388.37710963475644,
            "sha256": "cc21758ed0d374d7022ec1be971528d81a755ceb6dd836aac0bc64de1d25bf88",
            "sum": -1062.9958727225894,
            "top1": 72
          },
          "net_j": {
            "core": 4.082433837570317,
            "dram": 2.1675820039829317,
            "package-0": 31.603896471651325,
            "psys": 51.94530337567546,
            "uncore": 23.785907853002463
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.709174643,
          "watts": {
            "core": 6.1602979210591995,
            "dram": 1.7620136200440926,
            "package-0": 24.21745558274117,
            "psys": 40.54845318694562,
            "uncore": 13.91855776554485
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 450.3147795636656,
            "sha256": "4f4fcbcd85ba65dc12597e879daaf0bd341940cd416489da4580479054b6afca",
            "sum": 3555.901248254231,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 57,
          "gross_j": {
            "core": 5.058764,
            "dram": 1.65411,
            "package-0": 13.292752,
            "psys": 22.839175,
            "uncore": 4.836291
          },
          "net_j": {
            "core": 2.910493829249845,
            "dram": 1.37285204401004,
            "package-0": 10.030998353955686,
            "psys": 17.054411742900733,
            "uncore": 4.835178591683776
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.569567514,
          "token_latencies_s": [
            0.046172866,
            0.040434646,
            0.039540705,
            0.039966414,
            0.03973317,
            0.039857281,
            0.039970843,
            0.040040039,
            0.042041807,
            0.039726003,
            0.041269873,
            0.039839921,
            0.04138198,
            0.039562882
          ],
          "watts": {
            "core": 8.881763576143845,
            "dram": 2.904150885262743,
            "package-0": 23.338325436868224,
            "psys": 40.09915319713968,
            "uncore": 8.49116370074435
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 4.908063311129808,
    "compile_target": "NPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402443,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 4096,
      "NPUW_LLM_MIN_RESPONSE_LEN": 4096,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.025964446365833282,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 21.981327,
            "dram": 6.124497,
            "package-0": 62.326867,
            "psys": 110.383691,
            "uncore": 24.162231
          },
          "net_j": {
            "core": 11.094065291328032,
            "dram": 4.699104134520901,
            "package-0": 45.7965599384356,
            "psys": 81.0669726652864,
            "uncore": 24.158231171876977
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.358756,
            "dram": 0.274353,
            "package-0": 4.08312,
            "psys": 6.517379,
            "uncore": 0.0
          },
          "logits": {
            "l2": 939.2982638865852,
            "sha256": "d408c999cd01a994cd8f761c9d707a0e2e7e44d9d66fb15053e70ed4832f1885",
            "sum": -187687.95846249163,
            "top1": 285
          },
          "net_j": {
            "core": 2.534318931138132,
            "dram": 0.16641523314571613,
            "package-0": 2.8313637488971537,
            "psys": 4.297372723006728,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 61033,
          "npu_utilization": 0.2792228107354196,
          "seconds": 0.218581712,
          "watts": {
            "core": 15.366134564816658,
            "dram": 1.2551507511296278,
            "package-0": 18.680062310061878,
            "psys": 29.816671030557213,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1419.0084209097367,
          "decode_tokens_per_s": 17.887835209400944,
          "end_to_end_ms": 2886.5226869999997,
          "handoff_estimate_ms": 175.831959,
          "handoff_plus_first_decode_ms": 218.581712,
          "output_tokens_per_s_e2e": 5.543001644178659,
          "prefill_tokens_per_s": 1992.2225409376806,
          "steady_tpot_ms": 42.749753,
          "steady_tpot_p95_ms": 50.35086325,
          "ttft_ms": 2047.9639779999998
        },
        "prefill": {
          "gpu_active_fraction": 0.9901477832512315,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 203,
          "gross_j": {
            "core": 13.782923,
            "dram": 4.066274,
            "package-0": 47.835998,
            "psys": 85.236843,
            "uncore": 24.162231
          },
          "logits": {
            "l2": 371.2406598563232,
            "sha256": "bdb308b75faeea2b2633af4eb1c7004c83d886cd7bf96bf3bb48fc7235b4a04e",
            "sum": -41810.29560299311,
            "top1": 90
          },
          "net_j": {
            "core": 6.058500843598317,
            "dram": 3.054969548402806,
            "package-0": 36.1078818276143,
            "psys": 64.43687375913271,
            "uncore": 24.158231171876977
          },
          "npu_busy_delta_us": 9847,
          "npu_utilization": 0.004808190039366015,
          "seconds": 2.047963978,
          "watts": {
            "core": 6.730061245247157,
            "dram": 1.9855202746149083,
            "package-0": 23.357831736237696,
            "psys": 41.6202842997466,
            "uncore": 11.798171871946861
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1007.3695891492324,
            "sha256": "f7e5903d64b65e5e0eaea38dd2f8fdc85c5016fc0f999d164beb841f65c04032",
            "sum": -228969.46392723545,
            "top1": 83
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 4.839648,
            "dram": 1.78387,
            "package-0": 10.407749,
            "psys": 18.629469,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.5012455165915837,
            "dram": 1.4777193529723793,
            "package-0": 6.857314361924146,
            "psys": 12.332726183146963,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 609812,
          "npu_utilization": 0.9836042352390698,
          "seconds": 0.619976997,
          "token_latencies_s": [
            0.054577069,
            0.048075214,
            0.045824008,
            0.044201008,
            0.043082338,
            0.042682271,
            0.042702843,
            0.042517596,
            0.042668934,
            0.042728472,
            0.042929054,
            0.042771034,
            0.042558382,
            0.042633422
          ],
          "watts": {
            "core": 7.806173492594921,
            "dram": 2.8773164304997594,
            "package-0": 16.787314771938227,
            "psys": 30.04864549837484,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 20.939216000000002,
            "dram": 6.061569,
            "package-0": 61.011257,
            "psys": 106.939178,
            "uncore": 24.059997000000003
          },
          "net_j": {
            "core": 10.351644628989332,
            "dram": 4.675412489867736,
            "package-0": 44.93597466710354,
            "psys": 78.4294520433472,
            "uncore": 24.055617264842965
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.07692307692307693,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.491632,
            "dram": 0.206604,
            "package-0": 3.046379,
            "psys": 4.602405,
            "uncore": 0.023377
          },
          "logits": {
            "l2": 898.7402652849972,
            "sha256": "2c1f8e1f8b3933228503d4bde8c6fbdd7d411e07685d7be43122a51b53e1c935",
            "sum": -192974.0822976362,
            "top1": 16
          },
          "net_j": {
            "core": 2.0054836631892807,
            "dram": 0.14295600572954212,
            "package-0": 2.308252049441899,
            "psys": 3.2933270848602314,
            "uncore": 0.02312526469214614
          },
          "npu_busy_delta_us": 58612,
          "npu_utilization": 0.4547381849865347,
          "seconds": 0.128891749,
          "watts": {
            "core": 19.331198616910694,
            "dram": 1.6029264991974002,
            "package-0": 23.635174661180212,
            "psys": 35.707522286783465,
            "uncore": 0.18136925118457348
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1459.1746779203918,
          "decode_tokens_per_s": 19.842784979637216,
          "end_to_end_ms": 2807.066256,
          "handoff_estimate_ms": 85.7579085,
          "handoff_plus_first_decode_ms": 128.891749,
          "output_tokens_per_s_e2e": 5.69990108562653,
          "prefill_tokens_per_s": 1989.1532817151763,
          "steady_tpot_ms": 43.133840500000005,
          "steady_tpot_p95_ms": 51.68753355,
          "ttft_ms": 2051.123982
        },
        "prefill": {
          "gpu_active_fraction": 0.9901477832512315,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 203,
          "gross_j": {
            "core": 14.413537,
            "dram": 4.069325,
            "package-0": 48.354673,
            "psys": 84.738796,
            "uncore": 24.036498
          },
          "logits": {
            "l2": 369.66770690126174,
            "sha256": "080bfd9ddf05d67e276fff0c1594529574c430e3233356ed92f30ba9ca1508c0",
            "sum": -18263.070847075433,
            "top1": 36802
          },
          "net_j": {
            "core": 6.6771960768748135,
            "dram": 3.0564601078156697,
            "package-0": 36.60846036966127,
            "psys": 63.90673245146563,
            "uncore": 24.032492000150818
          },
          "npu_busy_delta_us": 7723,
          "npu_utilization": 0.003765252645756447,
          "seconds": 2.051123982,
          "watts": {
            "core": 7.027140790361058,
            "dram": 1.9839488181655907,
            "package-0": 23.574719726523092,
            "psys": 41.3133466058806,
            "uncore": 11.718695803343204
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 873.3404111901898,
            "sha256": "c166bc38f102527b2e6731e4c39fd240f13a426ac388431ea0b4312f82ce49bd",
            "sum": -176208.4787269421,
            "top1": 16
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 4.034047,
            "dram": 1.78564,
            "package-0": 9.610205,
            "psys": 17.597977,
            "uncore": 0.000122
          },
          "net_j": {
            "core": 1.6689648889252378,
            "dram": 1.4759963763225246,
            "package-0": 6.019262248000365,
            "psys": 11.229392507021341,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 613424,
          "npu_utilization": 0.9782688564051516,
          "seconds": 0.627050525,
          "token_latencies_s": [
            0.056709221,
            0.048983548,
            0.04656247,
            0.044554815,
            0.043137564,
            0.042737479,
            0.042900163,
            0.043371374,
            0.042844431,
            0.042893608,
            0.04320063,
            0.043130117,
            0.042918209,
            0.04307913
          ],
          "watts": {
            "core": 6.433368347789838,
            "dram": 2.8476812135672795,
            "package-0": 15.326045696237955,
            "psys": 28.06468745082384,
            "uncore": 0.00019456167427656646
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 23.035463999999997,
            "dram": 6.062302,
            "package-0": 62.961752000000004,
            "psys": 109.02706400000001,
            "uncore": 23.844848
          },
          "net_j": {
            "core": 12.44004145686111,
            "dram": 4.675117590930733,
            "package-0": 46.8745491043812,
            "psys": 80.49619676760318,
            "uncore": 23.840585124055476
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.07692307692307693,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.88659,
            "dram": 0.206787,
            "package-0": 3.42327,
            "psys": 4.981249,
            "uncore": 0.002746
          },
          "logits": {
            "l2": 855.424098241804,
            "sha256": "6e333ff040eb0721a0fb72967308ed20c3ecf95f25bba81031549aff4eb53398",
            "sum": -173123.48829822987,
            "top1": 7952
          },
          "net_j": {
            "core": 2.3964403006442345,
            "dram": 0.1426151353775177,
            "package-0": 2.6790677154135487,
            "psys": 3.6613963995603687,
            "uncore": 0.0024921927232513775
          },
          "npu_busy_delta_us": 58583,
          "npu_utilization": 0.45080274793066133,
          "seconds": 0.129952624,
          "watts": {
            "core": 22.212633428625498,
            "dram": 1.5912491309140477,
            "package-0": 26.34244615176066,
            "psys": 38.33126909388148,
            "uncore": 0.02113077762862257
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1458.0934344386096,
          "decode_tokens_per_s": 19.82939927041136,
          "end_to_end_ms": 2809.1478249999996,
          "handoff_estimate_ms": 87.09321399999999,
          "handoff_plus_first_decode_ms": 129.952624,
          "output_tokens_per_s_e2e": 5.695677478275819,
          "prefill_tokens_per_s": 1987.630646140281,
          "steady_tpot_ms": 42.859410000000004,
          "steady_tpot_p95_ms": 52.703164599999994,
          "ttft_ms": 2052.695257
        },
        "prefill": {
          "gpu_active_fraction": 0.9901477832512315,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 203,
          "gross_j": {
            "core": 15.602255,
            "dram": 4.066762,
            "package-0": 49.407527,
            "psys": 85.858179,
            "uncore": 23.842102
          },
          "logits": {
            "l2": 384.14249037604066,
            "sha256": "2bf92929ffc495c30c149948ed262893f16b023751064a74e2d7ac66c13a72fe",
            "sum": 25864.579894389957,
            "top1": 36802
          },
          "net_j": {
            "core": 7.859987609635066,
            "dram": 3.0531211969992422,
            "package-0": 37.65231611776977,
            "psys": 65.01015693265377,
            "uncore": 23.838092931332223
          },
          "npu_busy_delta_us": 7793,
          "npu_utilization": 0.0037964719670027475,
          "seconds": 2.052695257,
          "watts": {
            "core": 7.600862790905743,
            "dram": 1.9811815641565542,
            "package-0": 24.069586964510634,
            "psys": 41.82704602995047,
            "uncore": 11.615022696961393
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 852.712668795079,
            "sha256": "4502f07f983dd39c0355d7b49af58e80332f12949df77a2d9ec746487eb958b2",
            "sum": -171446.30799585953,
            "top1": 7952
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 4.546619,
            "dram": 1.788753,
            "package-0": 10.130955,
            "psys": 18.187636,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.183613546581808,
            "dram": 1.4793812585539732,
            "package-0": 6.543165271197879,
            "psys": 11.824643435389044,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 612505,
          "npu_utilization": 0.9776616995196411,
          "seconds": 0.626499944,
          "token_latencies_s": [
            0.060635632,
            0.048431836,
            0.045668249,
            0.044093972,
            0.04302695,
            0.042891686,
            0.042752533,
            0.042605508,
            0.042652026,
            0.042923348,
            0.042827134,
            0.0426689,
            0.042623286,
            0.042675365
          ],
          "watts": {
            "core": 7.257173833043471,
            "dram": 2.8551526893671997,
            "package-0": 16.170719721564733,
            "psys": 29.030546888604356,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.496285448782146,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402290,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.025952915661036968,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 15.143148,
            "dram": 4.665576,
            "package-0": 54.364118,
            "psys": 90.98621700000001,
            "uncore": 28.793262
          },
          "net_j": {
            "core": 6.569871967329371,
            "dram": 3.5431370600948524,
            "package-0": 41.34717313950183,
            "psys": 67.9004930244886,
            "uncore": 28.78882262189668
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 1.816523,
            "dram": 0.139892,
            "package-0": 2.443475,
            "psys": 3.472342,
            "uncore": 0.31671
          },
          "logits": {
            "l2": 386.3595092072282,
            "sha256": "a08f57e55d540bd7c46befedb70a9eeed4a0b576f2871d968449003d95edde09",
            "sum": 34339.569468914415,
            "top1": 3862
          },
          "net_j": {
            "core": 1.5665191241805552,
            "dram": 0.10716074484414066,
            "package-0": 2.0638900369063977,
            "psys": 2.7991430445503096,
            "uncore": 0.3165805440338294
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.066283137,
          "watts": {
            "core": 27.405507376634873,
            "dram": 2.1105217153497122,
            "package-0": 36.86420273077901,
            "psys": 52.386506691739704,
            "uncore": 4.778138367229058
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1802.0084721850667,
          "decode_tokens_per_s": 23.255637931745145,
          "end_to_end_ms": 2273.0192799999995,
          "handoff_estimate_ms": 25.405990000000003,
          "handoff_plus_first_decode_ms": 66.28313700000001,
          "output_tokens_per_s_e2e": 7.039095594472917,
          "prefill_tokens_per_s": 2506.1203420634615,
          "steady_tpot_ms": 40.877147,
          "steady_tpot_p95_ms": 43.18287125,
          "ttft_ms": 1628.014398
        },
        "prefill": {
          "gpu_active_fraction": 0.8571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 161,
          "gross_j": {
            "core": 9.233069,
            "dram": 2.865471,
            "package-0": 39.549703,
            "psys": 65.546158,
            "uncore": 23.605043
          },
          "logits": {
            "l2": 388.37710963475644,
            "sha256": "cc21758ed0d374d7022ec1be971528d81a755ceb6dd836aac0bc64de1d25bf88",
            "sum": -1062.9958727225894,
            "top1": 72
          },
          "net_j": {
            "core": 3.0925945473521095,
            "dram": 2.06154171918834,
            "package-0": 30.226520465946386,
            "psys": 49.011370393453014,
            "uncore": 23.601863365082707
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.628014398,
          "watts": {
            "core": 5.671368147199888,
            "dram": 1.7601017555619922,
            "package-0": 24.293214512467724,
            "psys": 40.26141174213375,
            "uncore": 14.499283930780077
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 450.3147795636656,
            "sha256": "4f4fcbcd85ba65dc12597e879daaf0bd341940cd416489da4580479054b6afca",
            "sum": 3555.901248254231,
            "top1": 42
          },
          "gpu_active_fraction": 0.9824561403508771,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 57,
          "gross_j": {
            "core": 4.093556,
            "dram": 1.660213,
            "package-0": 12.37094,
            "psys": 21.967717,
            "uncore": 4.871509
          },
          "net_j": {
            "core": 1.9107582957967066,
            "dram": 1.3744345960623716,
            "package-0": 9.056762636649045,
            "psys": 16.089979586485278,
            "uncore": 4.870378712780146
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.578721745,
          "token_latencies_s": [
            0.046900673,
            0.041089277,
            0.040752755,
            0.04089594,
            0.040669917,
            0.040930185,
            0.040825164,
            0.040834051,
            0.041152592,
            0.040858354,
            0.041180978,
            0.0406794,
            0.041161525,
            0.040757978
          ],
          "watts": {
            "core": 7.073444250829041,
            "dram": 2.868758629416975,
            "package-0": 21.37631790559382,
            "psys": 37.95903158952495,
            "uncore": 8.417705127012292
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 18.963147,
            "dram": 4.823778,
            "package-0": 58.784762,
            "psys": 96.56927,
            "uncore": 29.017687000000002
          },
          "net_j": {
            "core": 10.08583415514328,
            "dram": 3.6615336513034142,
            "package-0": 45.30619308765672,
            "psys": 72.66484965943805,
            "uncore": 29.013090186820357
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.931086,
            "dram": 0.135131,
            "package-0": 2.487115,
            "psys": 3.469595,
            "uncore": 0.258667
          },
          "logits": {
            "l2": 386.3595092072282,
            "sha256": "a08f57e55d540bd7c46befedb70a9eeed4a0b576f2871d968449003d95edde09",
            "sum": 34339.569468914415,
            "top1": 3862
          },
          "net_j": {
            "core": 1.6982450661889696,
            "dram": 0.10464676854431859,
            "package-0": 2.1335888117453394,
            "psys": 2.8426116265873365,
            "uncore": 0.2585464312770081
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.061732753,
          "watts": {
            "core": 31.28138477802861,
            "dram": 2.188967661947621,
            "package-0": 40.28841869404399,
            "psys": 56.20347111362424,
            "uncore": 4.190109584129513
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1740.2919459129414,
          "decode_tokens_per_s": 23.802358204040285,
          "end_to_end_ms": 2353.628085,
          "handoff_estimate_ms": 21.885813000000002,
          "handoff_plus_first_decode_ms": 61.732753,
          "output_tokens_per_s_e2e": 6.798015413722427,
          "prefill_tokens_per_s": 2367.3604687284605,
          "steady_tpot_ms": 39.84694,
          "steady_tpot_p95_ms": 43.49010445,
          "ttft_ms": 1723.438426
        },
        "prefill": {
          "gpu_active_fraction": 0.8362573099415205,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 171,
          "gross_j": {
            "core": 11.666108,
            "dram": 3.037406,
            "package-0": 42.719556,
            "psys": 70.018619,
            "uncore": 23.925781
          },
          "logits": {
            "l2": 388.37710963475644,
            "sha256": "cc21758ed0d374d7022ec1be971528d81a755ceb6dd836aac0bc64de1d25bf88",
            "sum": -1062.9958727225894,
            "top1": 72
          },
          "net_j": {
            "core": 5.165717316376147,
            "dram": 2.186355409759852,
            "package-0": 32.849906780964005,
            "psys": 52.51466549573151,
            "uncore": 23.922414994758377
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.723438426,
          "watts": {
            "core": 6.769088946842363,
            "dram": 1.7624105127153524,
            "package-0": 24.787399048046986,
            "psys": 40.62728203322536,
            "uncore": 13.882585324229042
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 450.3147795636656,
            "sha256": "4f4fcbcd85ba65dc12597e879daaf0bd341940cd416489da4580479054b6afca",
            "sum": 3555.901248254231,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 56,
          "gross_j": {
            "core": 5.365953,
            "dram": 1.651241,
            "package-0": 13.578091,
            "psys": 23.081056,
            "uncore": 4.833239
          },
          "net_j": {
            "core": 3.2218717725781625,
            "dram": 1.3705314729992435,
            "package-0": 10.322697494947377,
            "psys": 17.307572537119196,
            "uncore": 4.83212876078497
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.568456906,
          "token_latencies_s": [
            0.04644419,
            0.040632549,
            0.039936178,
            0.039789846,
            0.041899443,
            0.03980609,
            0.039836461,
            0.039904495,
            0.039776595,
            0.03963153,
            0.039857419,
            0.039720383,
            0.039661209,
            0.041537716
          ],
          "watts": {
            "core": 9.43950709959358,
            "dram": 2.904777798583029,
            "package-0": 23.885875704358142,
            "psys": 40.603000432191074,
            "uncore": 8.502384171932286
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 20.296885,
            "dram": 4.785206,
            "package-0": 59.998078,
            "psys": 98.63902999999999,
            "uncore": 28.968310000000002
          },
          "net_j": {
            "core": 11.486442752495272,
            "dram": 3.6317165499112005,
            "package-0": 46.6210398065094,
            "psys": 74.91467573297432,
            "uncore": 28.963747813474683
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.948421,
            "dram": 0.134827,
            "package-0": 2.496271,
            "psys": 3.514701,
            "uncore": 0.254577
          },
          "logits": {
            "l2": 386.3595092072282,
            "sha256": "a08f57e55d540bd7c46befedb70a9eeed4a0b576f2871d968449003d95edde09",
            "sum": 34339.569468914415,
            "top1": 3862
          },
          "net_j": {
            "core": 1.7227560862980686,
            "dram": 0.10528227455967387,
            "package-0": 2.1536402801429517,
            "psys": 2.907040883980436,
            "uncore": 0.25446014713386605
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.059830186,
          "watts": {
            "core": 32.56585229402429,
            "dram": 2.2534945821495524,
            "package-0": 41.722601363799875,
            "psys": 58.7446109560816,
            "uncore": 4.254992621951735
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1753.5006315521923,
          "decode_tokens_per_s": 24.010657466465695,
          "end_to_end_ms": 2335.8987879999995,
          "handoff_estimate_ms": 20.0771685,
          "handoff_plus_first_decode_ms": 59.830186,
          "output_tokens_per_s_e2e": 6.849611842000751,
          "prefill_tokens_per_s": 2384.324882993946,
          "steady_tpot_ms": 39.7530175,
          "steady_tpot_p95_ms": 43.07203225000001,
          "ttft_ms": 1711.176203
        },
        "prefill": {
          "gpu_active_fraction": 0.8402366863905325,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 169,
          "gross_j": {
            "core": 12.681852,
            "dram": 3.00103,
            "package-0": 43.607433,
            "psys": 71.210145,
            "uncore": 23.875489
          },
          "logits": {
            "l2": 388.37710963475644,
            "sha256": "cc21758ed0d374d7022ec1be971528d81a755ceb6dd836aac0bc64de1d25bf88",
            "sum": -1062.9958727225894,
            "top1": 72
          },
          "net_j": {
            "core": 6.227711439935663,
            "dram": 2.156034614265201,
            "package-0": 33.808006098065576,
            "psys": 53.8307317044288,
            "uncore": 23.87214694380481
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.711176203,
          "watts": {
            "core": 7.411190021089839,
            "dram": 1.7537819861792456,
            "package-0": 25.483894016027293,
            "psys": 41.61473545223209,
            "uncore": 13.952677087340259
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 450.3147795636656,
            "sha256": "4f4fcbcd85ba65dc12597e879daaf0bd341940cd416489da4580479054b6afca",
            "sum": 3555.901248254231,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 56,
          "gross_j": {
            "core": 5.666612,
            "dram": 1.649349,
            "package-0": 13.894374,
            "psys": 23.914184,
            "uncore": 4.838244
          },
          "net_j": {
            "core": 3.5359752262615407,
            "dram": 1.3703996610863258,
            "package-0": 10.659393428300875,
            "psys": 18.176903144565085,
            "uncore": 4.837140722536006
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.564892399,
          "token_latencies_s": [
            0.045324669,
            0.040117195,
            0.039953954,
            0.039629823,
            0.039756715,
            0.039680659,
            0.039402416,
            0.03974932,
            0.039635683,
            0.039542059,
            0.039846411,
            0.041859074,
            0.039652525,
            0.040717917
          ],
          "watts": {
            "core": 10.031312175613111,
            "dram": 2.9197578209934454,
            "package-0": 24.59649665068338,
            "psys": 42.33405165715462,
            "uncore": 8.564894851771585
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.156899374909699,
    "compile_target": "AUTO:GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402857,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.026081707328557968,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 34.007787,
            "dram": 8.636512999999999,
            "package-0": 124.026293,
            "psys": 203.921414,
            "uncore": 69.746404
          },
          "net_j": {
            "core": 15.122852288488216,
            "dram": 6.164040864845366,
            "package-0": 95.35298854822662,
            "psys": 153.06892909409794,
            "uncore": 69.73662508172869
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.245722,
            "dram": 0.152954,
            "package-0": 2.92498,
            "psys": 4.14507,
            "uncore": 0.342589
          },
          "logits": {
            "l2": 434.858175530361,
            "sha256": "9f8607463d2448c1a2e372ef90cbd5de694c07a24236a1d6b9f47ec97f4c074c",
            "sum": 63194.81769055477,
            "top1": 546
          },
          "net_j": {
            "core": 1.954787633673432,
            "dram": 0.11486400260843758,
            "package-0": 2.4832496054500868,
            "psys": 3.361655299433866,
            "uncore": 0.34243834957763514
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.077134974,
          "watts": {
            "core": 29.11418625745567,
            "dram": 1.9829396714387952,
            "package-0": 37.9202824389362,
            "psys": 53.73788030316831,
            "uncore": 4.441422382536876
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1636.1312634918727,
          "decode_tokens_per_s": 21.094972489912774,
          "end_to_end_ms": 5006.933235,
          "handoff_estimate_ms": 32.444108,
          "handoff_plus_first_decode_ms": 77.134974,
          "output_tokens_per_s_e2e": 3.195568874007564,
          "prefill_tokens_per_s": 1903.2262817325463,
          "steady_tpot_ms": 44.69086599999999,
          "steady_tpot_p95_ms": 47.949569,
          "ttft_ms": 4295.863334
        },
        "prefill": {
          "gpu_active_fraction": 0.8938679245283019,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 424,
          "gross_j": {
            "core": 26.235285,
            "dram": 6.716657,
            "package-0": 106.385836,
            "psys": 174.396465,
            "uncore": 63.859395
          },
          "logits": {
            "l2": 408.920230973876,
            "sha256": "b215d51e686e0c73b412e329e624a7ec911f718cf36cfb8e899b127b30d3aab8",
            "sum": -40438.595054788515,
            "top1": 198
          },
          "net_j": {
            "core": 10.03233305627045,
            "dram": 4.595318070313336,
            "package-0": 81.78462973706243,
            "psys": 130.76590012188592,
            "uncore": 63.85100485487998
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.295863334,
          "watts": {
            "core": 6.107104197742619,
            "dram": 1.563517383535088,
            "package-0": 24.764716130049962,
            "psys": 40.59637177461475,
            "uncore": 14.865322761685416
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 565.7297862327463,
            "sha256": "e36bc34d8a2d387ad8383ae4653516b9609d7b60cccdd5ffcf42ce7e972bfcad",
            "sum": 109975.08116664726,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 5.52678,
            "dram": 1.766902,
            "package-0": 14.715477,
            "psys": 25.379879,
            "uncore": 5.54442
          },
          "net_j": {
            "core": 3.1357315985443353,
            "dram": 1.4538587919235921,
            "package-0": 11.085109205714096,
            "psys": 18.94137367277815,
            "uncore": 5.543181877271076
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.633934927,
          "token_latencies_s": [
            0.051501117,
            0.046037197,
            0.045284153,
            0.045013893,
            0.04470893,
            0.044451329,
            0.044612654,
            0.044750672,
            0.044902173,
            0.044556705,
            0.04426657,
            0.044647506,
            0.044672802,
            0.044499663
          ],
          "watts": {
            "core": 8.718213438963902,
            "dram": 2.787197746559877,
            "package-0": 23.212914091417463,
            "psys": 40.035464081631204,
            "uncore": 8.746039638860283
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 35.521393,
            "dram": 8.999671,
            "package-0": 127.029032,
            "psys": 209.3271,
            "uncore": 70.442874
          },
          "net_j": {
            "core": 15.976562026294566,
            "dram": 6.440803272523199,
            "package-0": 97.35379628812373,
            "psys": 156.69767674481108,
            "uncore": 70.43275337699346
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.280878,
            "dram": 0.152221,
            "package-0": 2.919853,
            "psys": 4.096059,
            "uncore": 0.306701
          },
          "logits": {
            "l2": 434.858175530361,
            "sha256": "9f8607463d2448c1a2e372ef90cbd5de694c07a24236a1d6b9f47ec97f4c074c",
            "sum": 63194.81769055477,
            "top1": 546
          },
          "net_j": {
            "core": 1.995239470711877,
            "dram": 0.11482434943276595,
            "package-0": 2.486163361219168,
            "psys": 3.3269046862011815,
            "uncore": 0.306553091845923
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.075730897,
          "watts": {
            "core": 30.118196011860256,
            "dram": 2.010025049617463,
            "package-0": 38.555637337822624,
            "psys": 54.08702606546441,
            "uncore": 4.049879403910929
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1580.8902175759986,
          "decode_tokens_per_s": 21.236507844865102,
          "end_to_end_ms": 5181.8904999999995,
          "handoff_estimate_ms": 31.548724,
          "handoff_plus_first_decode_ms": 75.730897,
          "output_tokens_per_s_e2e": 3.087676206203122,
          "prefill_tokens_per_s": 1826.810630307582,
          "steady_tpot_ms": 44.182173000000006,
          "steady_tpot_p95_ms": 48.84928345,
          "ttft_ms": 4475.55968
        },
        "prefill": {
          "gpu_active_fraction": 0.8733031674208145,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 442,
          "gross_j": {
            "core": 27.806874,
            "dram": 7.082318,
            "package-0": 109.471888,
            "psys": 180.056301,
            "uncore": 64.575946
          },
          "logits": {
            "l2": 408.920230973876,
            "sha256": "b215d51e686e0c73b412e329e624a7ec911f718cf36cfb8e899b127b30d3aab8",
            "sum": -40438.595054788515,
            "top1": 198
          },
          "net_j": {
            "core": 10.926151132866629,
            "dram": 4.8722432738844255,
            "package-0": 83.84161108724595,
            "psys": 134.60066570578104,
            "uncore": 64.56720489436209
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.47555968,
          "watts": {
            "core": 6.213049537527338,
            "dram": 1.5824429806285143,
            "package-0": 24.45993257317038,
            "psys": 40.2310132975369,
            "uncore": 14.428574439208461
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 565.7297862327463,
            "sha256": "e36bc34d8a2d387ad8383ae4653516b9609d7b60cccdd5ffcf42ce7e972bfcad",
            "sum": 109975.08116664726,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 5.433641,
            "dram": 1.765132,
            "package-0": 14.637291,
            "psys": 25.17474,
            "uncore": 5.560227
          },
          "net_j": {
            "core": 3.05517142271606,
            "dram": 1.4537356492060076,
            "package-0": 11.026021839658608,
            "psys": 18.770106352828847,
            "uncore": 5.558995390785442
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.630599923,
          "token_latencies_s": [
            0.050734353,
            0.047834246,
            0.044262776,
            0.044190022,
            0.044211449,
            0.044067744,
            0.044131718,
            0.046623565,
            0.043943286,
            0.04389234,
            0.044337456,
            0.044174324,
            0.044100972,
            0.044071358
          ],
          "watts": {
            "core": 8.616621730859297,
            "dram": 2.799131328152731,
            "package-0": 23.211691702030226,
            "psys": 39.92188879477551,
            "uncore": 8.817360734121118
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 36.630277,
            "dram": 9.026283000000001,
            "package-0": 128.287086,
            "psys": 212.735784,
            "uncore": 70.520754
          },
          "net_j": {
            "core": 17.053544178739912,
            "dram": 6.463238587226493,
            "package-0": 98.56341319255087,
            "psys": 160.02045691479918,
            "uncore": 70.51061685771158
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.391046,
            "dram": 0.153137,
            "package-0": 3.039054,
            "psys": 4.273427,
            "uncore": 0.313232
          },
          "logits": {
            "l2": 434.858175530361,
            "sha256": "9f8607463d2448c1a2e372ef90cbd5de694c07a24236a1d6b9f47ec97f4c074c",
            "sum": 63194.81769055477,
            "top1": 546
          },
          "net_j": {
            "core": 2.11127555492739,
            "dram": 0.11650861656904325,
            "package-0": 2.6142739692125434,
            "psys": 3.5200739938866157,
            "uncore": 0.31308713043286174
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.074175101,
          "watts": {
            "core": 32.23515664643315,
            "dram": 2.064533757763269,
            "package-0": 40.97134967163712,
            "psys": 57.612688656804124,
            "uncore": 4.222872578225408
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1578.3140308760924,
          "decode_tokens_per_s": 21.482071330361432,
          "end_to_end_ms": 5190.3485869999995,
          "handoff_estimate_ms": 30.237168999999994,
          "handoff_plus_first_decode_ms": 74.175101,
          "output_tokens_per_s_e2e": 3.082644591554868,
          "prefill_tokens_per_s": 1820.0874304553242,
          "steady_tpot_ms": 43.937931999999996,
          "steady_tpot_p95_ms": 47.9535477,
          "ttft_ms": 4492.091898
        },
        "prefill": {
          "gpu_active_fraction": 0.8741573033707866,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 445,
          "gross_j": {
            "core": 28.49242,
            "dram": 7.113324,
            "package-0": 110.328758,
            "psys": 182.669272,
            "uncore": 64.676592
          },
          "logits": {
            "l2": 408.920230973876,
            "sha256": "b215d51e686e0c73b412e329e624a7ec911f718cf36cfb8e899b127b30d3aab8",
            "sum": -40438.595054788515,
            "top1": 198
          },
          "net_j": {
            "core": 11.549341626429264,
            "dram": 4.895085504483314,
            "package-0": 84.60380570458587,
            "psys": 137.04572866170543,
            "uncore": 64.66781860569336
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.492091898,
          "watts": {
            "core": 6.342795438509526,
            "dram": 1.58352147763652,
            "package-0": 24.560663607332103,
            "psys": 40.66463379374079,
            "uncore": 14.397878197637889
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 565.7297862327463,
            "sha256": "e36bc34d8a2d387ad8383ae4653516b9609d7b60cccdd5ffcf42ce7e972bfcad",
            "sum": 109975.08116664726,
            "top1": 42
          },
          "gpu_active_fraction": 0.9838709677419355,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 5.746811,
            "dram": 1.759822,
            "package-0": 14.919274,
            "psys": 25.793085,
            "uncore": 5.53093
          },
          "net_j": {
            "core": 3.392926997383257,
            "dram": 1.4516444661741361,
            "package-0": 11.34533351875246,
            "psys": 19.454654259207153,
            "uncore": 5.529711121585361
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.624081588,
          "token_latencies_s": [
            0.050485663,
            0.043985239,
            0.044190389,
            0.043998546,
            0.046590101,
            0.043720321,
            0.043863455,
            0.043949691,
            0.043804734,
            0.043866955,
            0.04387884,
            0.043837604,
            0.043961435,
            0.043926173
          ],
          "watts": {
            "core": 9.208429010727361,
            "dram": 2.8198588675556313,
            "package-0": 23.90596724350086,
            "psys": 41.32966826125946,
            "uncore": 8.862511098468746
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 7.078193558380008,
    "compile_target": "NPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402928,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 8192,
      "NPUW_LLM_MIN_RESPONSE_LEN": 8192,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.025981368497014046,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 41.879715,
            "dram": 13.022060999999999,
            "package-0": 136.576677,
            "psys": 243.282398,
            "uncore": 61.252041000000006
          },
          "net_j": {
            "core": 18.448433684824945,
            "dram": 9.954367569880533,
            "package-0": 101.00058033308278,
            "psys": 180.18771974942854,
            "uncore": 61.2412469156494
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.038461538461538464,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 4.216603,
            "dram": 0.339415,
            "package-0": 5.105456,
            "psys": 8.046245,
            "uncore": 0.005799
          },
          "logits": {
            "l2": 854.470127280065,
            "sha256": "6266d336e326e5bcea5486fb86eccb6fe7396351155b33ffd895286994487164",
            "sum": -172627.30543726124,
            "top1": 1506
          },
          "net_j": {
            "core": 3.2114983896765588,
            "dram": 0.20782369826119576,
            "package-0": 3.5793892733770414,
            "psys": 5.339745464464475,
            "uncore": 0.0052785409150939485
          },
          "npu_busy_delta_us": 68633,
          "npu_utilization": 0.25755228363331284,
          "seconds": 0.266481815,
          "watts": {
            "core": 15.823229814011887,
            "dram": 1.2736891633674892,
            "package-0": 19.158740719324506,
            "psys": 30.19434928420913,
            "uncore": 0.021761334821289776
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1318.6744538164178,
          "decode_tokens_per_s": 15.755154612359009,
          "end_to_end_ms": 6212.2990070000005,
          "handoff_estimate_ms": 219.001846,
          "handoff_plus_first_decode_ms": 266.481815,
          "output_tokens_per_s_e2e": 2.575536042610191,
          "prefill_tokens_per_s": 1554.3047645783913,
          "steady_tpot_ms": 47.479969,
          "steady_tpot_p95_ms": 55.725221499999996,
          "ttft_ms": 5260.229645
        },
        "prefill": {
          "gpu_active_fraction": 0.9942196531791907,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 519,
          "gross_j": {
            "core": 32.901466,
            "dram": 10.709385,
            "package-0": 120.433529,
            "psys": 215.489011,
            "uncore": 61.246242
          },
          "logits": {
            "l2": 382.2709395892454,
            "sha256": "725790162612ec31b6bdfcc457776f6461f44d6aa1009d0b7216add0abf45935",
            "sum": 21197.831756433472,
            "top1": 36802
          },
          "net_j": {
            "core": 13.061158819753803,
            "dram": 8.11183264382666,
            "package-0": 90.30966694348331,
            "psys": 162.0639429854989,
            "uncore": 61.235968374734306
          },
          "npu_busy_delta_us": 9715,
          "npu_utilization": 0.0018468775425488102,
          "seconds": 5.260229645,
          "watts": {
            "core": 6.254758484028125,
            "dram": 2.035915867319515,
            "package-0": 22.89510860319103,
            "psys": 40.96570407431328,
            "uncore": 11.643263913052984
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 846.521889761484,
            "sha256": "0dd3730743829ee15aaf4b29d9da089dfab4f667b1d4f80f5652bc78e4e652bd",
            "sum": -167734.4824058395,
            "top1": 1506
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 4.761646,
            "dram": 1.973261,
            "package-0": 11.037692,
            "psys": 19.747142,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.175776475394582,
            "dram": 1.6347112277926783,
            "package-0": 7.111524116222437,
            "psys": 12.78403129946518,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 670982,
          "npu_utilization": 0.9786963064543527,
          "seconds": 0.685587547,
          "token_latencies_s": [
            0.060509241,
            0.053149211,
            0.04993172,
            0.047973488,
            0.047831694,
            0.047155661,
            0.047147659,
            0.047046709,
            0.046944158,
            0.047338623,
            0.048452045,
            0.047373583,
            0.047586355,
            0.047104668
          ],
          "watts": {
            "core": 6.945350773706512,
            "dram": 2.878204262365343,
            "package-0": 16.099609813945467,
            "psys": 28.803239041329903,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 42.762159,
            "dram": 12.936490000000001,
            "package-0": 136.692398,
            "psys": 239.878658,
            "uncore": 60.60037
          },
          "net_j": {
            "core": 19.62691710141337,
            "dram": 9.907554935721011,
            "package-0": 101.5657828087265,
            "psys": 177.58114109489182,
            "uncore": 60.59006664014917
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 16,
          "gross_j": {
            "core": 3.144035,
            "dram": 0.268737,
            "package-0": 3.838186,
            "psys": 5.772202,
            "uncore": 0.0
          },
          "logits": {
            "l2": 825.823560961018,
            "sha256": "dc26d19b3ff8d298bcd6d704495a510b3d83d3dd2060440d27fe5c4ef784ad90",
            "sum": -153551.2782026939,
            "top1": 7952
          },
          "net_j": {
            "core": 2.499851326154392,
            "dram": 0.18439854673895054,
            "package-0": 2.8601114191131716,
            "psys": 4.0375737870206,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 65432,
          "npu_utilization": 0.38311060033340905,
          "seconds": 0.170791411,
          "watts": {
            "core": 18.408624775633477,
            "dram": 1.573480764790918,
            "package-0": 22.472945082700907,
            "psys": 33.79679321227693,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1335.5482612176588,
          "decode_tokens_per_s": 17.47525852897468,
          "end_to_end_ms": 6133.810539,
          "handoff_estimate_ms": 123.2905085,
          "handoff_plus_first_decode_ms": 170.791411,
          "output_tokens_per_s_e2e": 2.60849269769074,
          "prefill_tokens_per_s": 1549.8191785152774,
          "steady_tpot_ms": 47.500902499999995,
          "steady_tpot_p95_ms": 56.091384250000004,
          "ttft_ms": 5275.4541389999995
        },
        "prefill": {
          "gpu_active_fraction": 0.9923224568138196,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 521,
          "gross_j": {
            "core": 34.969088,
            "dram": 10.702182,
            "package-0": 121.937248,
            "psys": 215.072203,
            "uncore": 60.60037
          },
          "logits": {
            "l2": 611.459166441154,
            "sha256": "bd78e88b7ae6c11ade7484e9b0e711a1ea87e31cb1727a506e055cb563fe9aa9",
            "sum": 160544.48878543824,
            "top1": 70047
          },
          "net_j": {
            "core": 15.071357730267412,
            "dram": 8.097111641205075,
            "package-0": 91.72619952994485,
            "psys": 161.49250872726898,
            "uncore": 60.59006664014917
          },
          "npu_busy_delta_us": 9630,
          "npu_utilization": 0.0018254352604087723,
          "seconds": 5.275454139,
          "watts": {
            "core": 6.628640317708959,
            "dram": 2.028675014134172,
            "package-0": 23.114076018318695,
            "psys": 40.768471743509174,
            "uncore": 11.487232834041324
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 824.9906646159244,
            "sha256": "5240a7a1fab300f6d868051f3c76151567b4ee8890b9bf6f9032d2a22ee47201",
            "sum": -152912.27035018802,
            "top1": 7952
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 4.649036,
            "dram": 1.965571,
            "package-0": 10.916964,
            "psys": 19.034253,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.055708044991565,
            "dram": 1.6260447477769857,
            "package-0": 6.979471859668472,
            "psys": 12.05105858060223,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 672336,
          "npu_utilization": 0.9778508370210225,
          "seconds": 0.687564989,
          "token_latencies_s": [
            0.060806988,
            0.053552213,
            0.050631662,
            0.048428059,
            0.047484321,
            0.047619957,
            0.047401296,
            0.047517484,
            0.047354344,
            0.047559459,
            0.047451612,
            0.047266378,
            0.047218905,
            0.047243619
          ],
          "watts": {
            "core": 6.761595011929847,
            "dram": 2.858742128302289,
            "package-0": 15.87771945147719,
            "psys": 27.683569269115303,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 42.853589,
            "dram": 12.872036999999999,
            "package-0": 136.347795,
            "psys": 240.07922000000002,
            "uncore": 60.456632
          },
          "net_j": {
            "core": 19.970343132448978,
            "dram": 9.876094009796086,
            "package-0": 101.60378949358383,
            "psys": 178.46026643441846,
            "uncore": 60.446119090590884
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 16,
          "gross_j": {
            "core": 3.050591,
            "dram": 0.258605,
            "package-0": 3.713125,
            "psys": 5.865219,
            "uncore": 0.002808
          },
          "logits": {
            "l2": 818.7195203975049,
            "sha256": "b893964f6d8b73b805ad2828c353b169a2e47afcb291c2a885575f97ed567554",
            "sum": -147686.32956599817,
            "top1": 7952
          },
          "net_j": {
            "core": 2.444403295540639,
            "dram": 0.17924109269268468,
            "package-0": 2.7927403192793614,
            "psys": 4.232904588429796,
            "uncore": 0.002494106406533452
          },
          "npu_busy_delta_us": 65427,
          "npu_utilization": 0.4070929407052678,
          "seconds": 0.160717599,
          "watts": {
            "core": 18.98106379749986,
            "dram": 1.609064605301875,
            "package-0": 23.103412588934955,
            "psys": 36.4939436408579,
            "uncore": 0.01747163980467379
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1350.2556529500794,
          "decode_tokens_per_s": 17.752097188973398,
          "end_to_end_ms": 6066.999225,
          "handoff_estimate_ms": 113.50485349999998,
          "handoff_plus_first_decode_ms": 160.71759899999998,
          "output_tokens_per_s_e2e": 2.6372180721681238,
          "prefill_tokens_per_s": 1565.6750600848916,
          "steady_tpot_ms": 47.2127455,
          "steady_tpot_p95_ms": 55.732401749999994,
          "ttft_ms": 5222.028637
        },
        "prefill": {
          "gpu_active_fraction": 0.9941972920696325,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 517,
          "gross_j": {
            "core": 34.930513,
            "dram": 10.64664,
            "package-0": 121.491327,
            "psys": 214.722046,
            "uncore": 60.453824
          },
          "logits": {
            "l2": 966.8558789071615,
            "sha256": "bb371fe043ec74e5e598b48b757dabc067e49e45408d2d4baa79743863d3df6c",
            "sum": 294894.52650701255,
            "top1": 128009
          },
          "net_j": {
            "core": 15.234290733239558,
            "dram": 8.067951671642046,
            "package-0": 91.58623141815724,
            "psys": 161.68496322347465,
            "uncore": 60.44362498418435
          },
          "npu_busy_delta_us": 7442,
          "npu_utilization": 0.0014251166581643545,
          "seconds": 5.222028637,
          "watts": {
            "core": 6.689069598834526,
            "dram": 2.0387938749635777,
            "package-0": 23.265159087636768,
            "psys": 41.118511775024565,
            "uncore": 11.576693312568672
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 818.2795658820971,
            "sha256": "3a7e9eb144765f50cf35ea63cec5d07510d7a98b269c3dd4c9b6806d6afb579b",
            "sum": -147559.66908771917,
            "top1": 7952
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 4.872485,
            "dram": 1.966792,
            "package-0": 11.143343,
            "psys": 19.491955,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.291649103668782,
            "dram": 1.6289012454613547,
            "package-0": 7.224817756147224,
            "psys": 12.542398622513998,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 671696,
          "npu_utilization": 0.9816486165177716,
          "seconds": 0.684252989,
          "token_latencies_s": [
            0.060789379,
            0.053009414,
            0.050029215,
            0.048535803,
            0.047775461,
            0.047343773,
            0.047067618,
            0.046939714,
            0.046996144,
            0.047171086,
            0.047254405,
            0.047109545,
            0.047063552,
            0.047137655
          ],
          "watts": {
            "core": 7.120882302788158,
            "dram": 2.8743637683985335,
            "package-0": 16.28541369806132,
            "psys": 28.486474028394785,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.4913084590807557,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 402771,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_1b_ov_fp16/openvino_model.xml",
      "key": "llama32_1b_fp16",
      "size_b": 1.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.02621446456760168,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 32.331826,
            "dram": 8.649026,
            "package-0": 122.510796,
            "psys": 201.54513799999998,
            "uncore": 69.86578800000001
          },
          "net_j": {
            "core": 13.418726848390946,
            "dram": 6.1728664921111465,
            "package-0": 93.79472901936649,
            "psys": 150.6168131832479,
            "uncore": 69.85599449773558
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 1.910822,
            "dram": 0.155578,
            "package-0": 2.633355,
            "psys": 3.814504,
            "uncore": 0.374877
          },
          "logits": {
            "l2": 434.858175530361,
            "sha256": "9f8607463d2448c1a2e372ef90cbd5de694c07a24236a1d6b9f47ec97f4c074c",
            "sum": 63194.81769055477,
            "top1": 546
          },
          "net_j": {
            "core": 1.6116990326341636,
            "dram": 0.11641592646918844,
            "package-0": 2.1791917189074645,
            "psys": 3.009039410614499,
            "uncore": 0.3747221093903355
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.079306005,
          "watts": {
            "core": 24.094291472631866,
            "dram": 1.9617429978978262,
            "package-0": 33.20498870167524,
            "psys": 48.09855193184929,
            "uncore": 4.726968657669744
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1633.6948187509818,
          "decode_tokens_per_s": 21.039996389312194,
          "end_to_end_ms": 5014.400429,
          "handoff_estimate_ms": 34.614231,
          "handoff_plus_first_decode_ms": 79.306005,
          "output_tokens_per_s_e2e": 3.1908101928730113,
          "prefill_tokens_per_s": 1900.7444322758727,
          "steady_tpot_ms": 44.691774,
          "steady_tpot_p95_ms": 48.0945765,
          "ttft_ms": 4301.4725499999995
        },
        "prefill": {
          "gpu_active_fraction": 0.8941176470588236,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 425,
          "gross_j": {
            "core": 24.506651,
            "dram": 6.732283,
            "package-0": 104.753883,
            "psys": 172.209276,
            "uncore": 63.932515
          },
          "logits": {
            "l2": 408.920230973876,
            "sha256": "b215d51e686e0c73b412e329e624a7ec911f718cf36cfb8e899b127b30d3aab8",
            "sum": -40438.595054788515,
            "top1": 198
          },
          "net_j": {
            "core": 8.282542457416227,
            "dram": 4.608174185114325,
            "package-0": 80.12055432788401,
            "psys": 128.52174160410203,
            "uncore": 63.924113899658046
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.30147255,
          "watts": {
            "core": 5.697270112766383,
            "dram": 1.5651112315014077,
            "package-0": 24.353028360020573,
            "psys": 40.034958725936775,
            "uncore": 14.862936879604174
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 565.7297862327463,
            "sha256": "e36bc34d8a2d387ad8383ae4653516b9609d7b60cccdd5ffcf42ce7e972bfcad",
            "sum": 109975.08116664726,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 5.914353,
            "dram": 1.761165,
            "package-0": 15.123558,
            "psys": 25.521358,
            "uncore": 5.558396
          },
          "net_j": {
            "core": 3.524485358340555,
            "dram": 1.4482763805276333,
            "package-0": 11.494982972575015,
            "psys": 19.086032168531354,
            "uncore": 5.5571584886871985
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.633621874,
          "token_latencies_s": [
            0.051816951,
            0.046090221,
            0.045064518,
            0.044570015,
            0.044172358,
            0.044642264,
            0.044328,
            0.044415373,
            0.044644226,
            0.044833058,
            0.044665584,
            0.044875997,
            0.044742084,
            0.044717964
          ],
          "watts": {
            "core": 9.334199532385462,
            "dram": 2.7795205188891567,
            "package-0": 23.86842787564496,
            "psys": 40.278530535705585,
            "uncore": 8.772418106260012
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.808863,
            "dram": 8.978554,
            "package-0": 129.486546,
            "psys": 212.124396,
            "uncore": 70.544985
          },
          "net_j": {
            "core": 18.13163891029228,
            "dram": 6.402352969803206,
            "package-0": 99.61029566030305,
            "psys": 159.13847064215634,
            "uncore": 70.5347958217413
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 1.915644,
            "dram": 0.151672,
            "package-0": 2.554986,
            "psys": 3.702872,
            "uncore": 0.302245
          },
          "logits": {
            "l2": 434.858175530361,
            "sha256": "9f8607463d2448c1a2e372ef90cbd5de694c07a24236a1d6b9f47ec97f4c074c",
            "sum": 63194.81769055477,
            "top1": 546
          },
          "net_j": {
            "core": 1.630622994058136,
            "dram": 0.11435619743620133,
            "package-0": 2.1222339569897573,
            "psys": 2.9353805247084037,
            "uncore": 0.3020974116092913
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.075567174,
          "watts": {
            "core": 25.350213572893434,
            "dram": 2.0071148882714605,
            "package-0": 33.8107919716569,
            "psys": 49.00106493330027,
            "uncore": 3.99968642469017
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1570.2536063848963,
          "decode_tokens_per_s": 21.04278621303582,
          "end_to_end_ms": 5216.99168,
          "handoff_estimate_ms": 31.057063999999997,
          "handoff_plus_first_decode_ms": 75.567174,
          "output_tokens_per_s_e2e": 3.0669015749705006,
          "prefill_tokens_per_s": 1815.2115043771373,
          "steady_tpot_ms": 44.510110000000005,
          "steady_tpot_p95_ms": 49.04146014999999,
          "ttft_ms": 4504.1583200000005
        },
        "prefill": {
          "gpu_active_fraction": 0.8677130044843049,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 446,
          "gross_j": {
            "core": 30.426741,
            "dram": 7.063519,
            "package-0": 112.203692,
            "psys": 183.083149,
            "uncore": 64.674151
          },
          "logits": {
            "l2": 408.920230973876,
            "sha256": "b215d51e686e0c73b412e329e624a7ec911f718cf36cfb8e899b127b30d3aab8",
            "sum": -40438.595054788515,
            "top1": 198
          },
          "net_j": {
            "core": 13.438151016632144,
            "dram": 4.83932198840855,
            "package-0": 86.4096386849775,
            "psys": 137.33705408895148,
            "uncore": 64.66535403906103
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.50415832,
          "watts": {
            "core": 6.755255663393288,
            "dram": 1.5682217404826924,
            "package-0": 24.911134118393957,
            "psys": 40.647582965067706,
            "uncore": 14.358765035594928
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 565.7297862327463,
            "sha256": "e36bc34d8a2d387ad8383ae4653516b9609d7b60cccdd5ffcf42ce7e972bfcad",
            "sum": 109975.08116664726,
            "top1": 42
          },
          "gpu_active_fraction": 0.984375,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 64,
          "gross_j": {
            "core": 5.466478,
            "dram": 1.763363,
            "package-0": 14.727868,
            "psys": 25.338375,
            "uncore": 5.568589
          },
          "net_j": {
            "core": 3.062864899602003,
            "dram": 1.4486747839584553,
            "package-0": 11.078423018335801,
            "psys": 18.866036028496463,
            "uncore": 5.567344371070979
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.637266186,
          "token_latencies_s": [
            0.051881342,
            0.046605175,
            0.04463742,
            0.047512293,
            0.044406685,
            0.046715937,
            0.044653599,
            0.044435373,
            0.044415668,
            0.044457714,
            0.044289617,
            0.044246308,
            0.044562506,
            0.044420155
          ],
          "watts": {
            "core": 8.57801358379307,
            "dram": 2.76707447961157,
            "package-0": 23.111014397992868,
            "psys": 39.76105363293197,
            "uncore": 8.738246469584377
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.494595000000004,
            "dram": 8.947732,
            "package-0": 129.174597,
            "psys": 212.203011,
            "uncore": 70.579043
          },
          "net_j": {
            "core": 17.921079053578353,
            "dram": 6.385108750101654,
            "package-0": 99.45580842609655,
            "psys": 159.4963461676301,
            "uncore": 70.56890752346031
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.299616,
            "dram": 0.1521,
            "package-0": 3.002068,
            "psys": 4.203603,
            "uncore": 0.323913
          },
          "logits": {
            "l2": 434.858175530361,
            "sha256": "9f8607463d2448c1a2e372ef90cbd5de694c07a24236a1d6b9f47ec97f4c074c",
            "sum": 63194.81769055477,
            "top1": 546
          },
          "net_j": {
            "core": 2.000508289390609,
            "dram": 0.11293992392937363,
            "package-0": 2.5479278834896677,
            "psys": 3.3981794933075946,
            "uncore": 0.3237581172905256
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.07930196,
          "watts": {
            "core": 28.998224003542912,
            "dram": 1.9179853814458054,
            "package-0": 37.85616395862095,
            "psys": 53.00755492045846,
            "uncore": 4.084552260751185
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1578.5734241658304,
          "decode_tokens_per_s": 21.289175594032894,
          "end_to_end_ms": 5189.495702,
          "handoff_estimate_ms": 35.1541915,
          "handoff_plus_first_decode_ms": 79.30196000000001,
          "output_tokens_per_s_e2e": 3.0831512190738875,
          "prefill_tokens_per_s": 1823.0010919172291,
          "steady_tpot_ms": 44.147768500000005,
          "steady_tpot_p95_ms": 46.964878549999995,
          "ttft_ms": 4484.912289000001
        },
        "prefill": {
          "gpu_active_fraction": 0.8738738738738738,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 444,
          "gross_j": {
            "core": 29.68596,
            "dram": 7.041608,
            "package-0": 111.49648,
            "psys": 182.702902,
            "uncore": 64.729814
          },
          "logits": {
            "l2": 408.920230973876,
            "sha256": "b215d51e686e0c73b412e329e624a7ec911f718cf36cfb8e899b127b30d3aab8",
            "sum": -40438.595054788515,
            "top1": 198
          },
          "net_j": {
            "core": 12.769961365833694,
            "dram": 4.826914864958608,
            "package-0": 85.81264331468881,
            "psys": 137.15227774034435,
            "uncore": 64.72105462801127
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.484912289,
          "watts": {
            "core": 6.619072589849705,
            "dram": 1.570065933568138,
            "package-0": 24.860347943361976,
            "psys": 40.737229677402944,
            "uncore": 14.432793737964671
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 565.7297862327463,
            "sha256": "e36bc34d8a2d387ad8383ae4653516b9609d7b60cccdd5ffcf42ce7e972bfcad",
            "sum": 109975.08116664726,
            "top1": 42
          },
          "gpu_active_fraction": 0.9838709677419355,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 5.509019,
            "dram": 1.754024,
            "package-0": 14.676049,
            "psys": 25.296506,
            "uncore": 5.525316
          },
          "net_j": {
            "core": 3.1506093983540486,
            "dram": 1.4452539612136726,
            "package-0": 11.095237227918076,
            "psys": 18.945888933978146,
            "uncore": 5.5240947781585135
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.625281453,
          "token_latencies_s": [
            0.050768492,
            0.044916779,
            0.044437505,
            0.044250153,
            0.044552905,
            0.04395325,
            0.043859333,
            0.044011696,
            0.043881639,
            0.04417409,
            0.044045581,
            0.04429722,
            0.043991104,
            0.044121447
          ],
          "watts": {
            "core": 8.810462830088133,
            "dram": 2.805175160057082,
            "package-0": 23.471108777633933,
            "psys": 40.456191173800896,
            "uncore": 8.83652629306438
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.5536438301205635,
    "compile_target": "AUTO:GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395389,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.043110777623951435,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 15.565939,
            "dram": 5.705856000000001,
            "package-0": 49.701655,
            "psys": 84.284452,
            "uncore": 22.461978000000002
          },
          "net_j": {
            "core": 10.873495316169034,
            "dram": 4.768363000401306,
            "package-0": 41.495562237781044,
            "psys": 65.31244432195557,
            "uncore": 22.45792078364072
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.392267,
            "dram": 0.300597,
            "package-0": 3.819387,
            "psys": 5.67479,
            "uncore": 0.807432
          },
          "logits": {
            "l2": 534.553078595872,
            "sha256": "3565942ee06dde1770dddad80ce01950d2e4f39473f435d47f4fdea8aaa08c44",
            "sum": -125913.45951230347,
            "top1": 546
          },
          "net_j": {
            "core": 2.1297754015669206,
            "dram": 0.24815438106513033,
            "package-0": 3.3603446435691176,
            "psys": 4.613510854966381,
            "uncore": 0.8072050425170565
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.106855062,
          "watts": {
            "core": 22.387961367707597,
            "dram": 2.813128310196479,
            "package-0": 35.743622515515455,
            "psys": 53.107357702904146,
            "uncore": 7.556328964555746
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 536.0696183352625,
          "decode_tokens_per_s": 11.257175278493813,
          "end_to_end_ms": 1910.199655,
          "handoff_estimate_ms": 19.85353849999999,
          "handoff_plus_first_decode_ms": 106.855062,
          "output_tokens_per_s_e2e": 8.376087786488476,
          "prefill_tokens_per_s": 1744.8013915164981,
          "steady_tpot_ms": 87.0015235,
          "steady_tpot_p95_ms": 89.6039883,
          "ttft_ms": 577.716183
        },
        "prefill": {
          "gpu_active_fraction": 0.896551724137931,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 3.293998,
            "dram": 1.395382,
            "package-0": 17.105181,
            "psys": 27.834462,
            "uncore": 10.765536
          },
          "logits": {
            "l2": 878.279080814622,
            "sha256": "952e315a56fbaeb7978e914a4c86da9699720487343b981aaec4300879317855",
            "sum": -272259.80647715216,
            "top1": 198
          },
          "net_j": {
            "core": 1.87482663387111,
            "dram": 1.1118488751249571,
            "package-0": 14.623350069121175,
            "psys": 22.096613691359966,
            "uncore": 10.764308945255983
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.577716183,
          "watts": {
            "core": 5.7017582282267485,
            "dram": 2.415341721524875,
            "package-0": 29.60827739180711,
            "psys": 48.180166696143935,
            "uncore": 18.634645032957298
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 534.7418300276124,
            "sha256": "a801c56c0f265ea4dad1913c86c55ea9541f3f9ec2db193cbf0ca5420b0121b9",
            "sum": -97346.80173273385,
            "top1": 42
          },
          "gpu_active_fraction": 0.9918032786885246,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 122,
          "gross_j": {
            "core": 9.879674,
            "dram": 4.009877,
            "package-0": 28.777087,
            "psys": 50.7752,
            "uncore": 10.88901
          },
          "net_j": {
            "core": 6.868893280731003,
            "dram": 3.408359744211218,
            "package-0": 23.51186752509075,
            "psys": 38.60231977562922,
            "uncore": 10.88640679586768
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.22562841,
          "token_latencies_s": [
            0.093123425,
            0.087629737,
            0.086805589,
            0.086817869,
            0.087708907,
            0.086866376,
            0.086873926,
            0.086878485,
            0.087124562,
            0.087206075,
            0.086816568,
            0.087515447,
            0.087364025,
            0.086851441
          ],
          "watts": {
            "core": 8.060904854514591,
            "dram": 3.271690642353828,
            "package-0": 23.479454918966837,
            "psys": 41.42789085641382,
            "uncore": 8.884430151223405
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.063496,
            "dram": 5.754318,
            "package-0": 50.259697,
            "psys": 84.860806,
            "uncore": 22.399052
          },
          "net_j": {
            "core": 11.299604254683455,
            "dram": 4.802550549392244,
            "package-0": 41.928656671068275,
            "psys": 65.59992683783813,
            "uncore": 22.3949330076742
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.546929,
            "dram": 0.295104,
            "package-0": 3.85137,
            "psys": 5.615343,
            "uncore": 0.704527
          },
          "logits": {
            "l2": 534.553078595872,
            "sha256": "3565942ee06dde1770dddad80ce01950d2e4f39473f435d47f4fdea8aaa08c44",
            "sum": -125913.45951230347,
            "top1": 546
          },
          "net_j": {
            "core": 2.3090300484081165,
            "dram": 0.24757469187047199,
            "package-0": 3.435334987896495,
            "psys": 4.653494325768257,
            "uncore": 0.7043213059984756
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.096843889,
          "watts": {
            "core": 26.299325918231144,
            "dram": 3.047213438526823,
            "package-0": 39.76884901844452,
            "psys": 57.98345211023072,
            "uncore": 7.27487306917218
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 528.0297347486922,
          "decode_tokens_per_s": 11.351113110507946,
          "end_to_end_ms": 1939.284727,
          "handoff_estimate_ms": 9.935859500000005,
          "handoff_plus_first_decode_ms": 96.843889,
          "output_tokens_per_s_e2e": 8.250464605448316,
          "prefill_tokens_per_s": 1631.5209299832452,
          "steady_tpot_ms": 86.9080295,
          "steady_tpot_p95_ms": 89.6742428,
          "ttft_ms": 617.828421
        },
        "prefill": {
          "gpu_active_fraction": 0.8548387096774194,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 3.793508,
            "dram": 1.447872,
            "package-0": 17.802566,
            "psys": 29.081834,
            "uncore": 10.830417
          },
          "logits": {
            "l2": 878.279080814622,
            "sha256": "952e315a56fbaeb7978e914a4c86da9699720487343b981aaec4300879317855",
            "sum": -272259.80647715216,
            "top1": 198
          },
          "net_j": {
            "core": 2.2758001184054404,
            "dram": 1.144652481493378,
            "package-0": 15.148415518614183,
            "psys": 22.945592944650176,
            "uncore": 10.829104747864289
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.617828421,
          "watts": {
            "core": 6.140067162756826,
            "dram": 2.343485587238791,
            "package-0": 28.814741107547718,
            "psys": 47.07105243382774,
            "uncore": 17.529813507883283
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 534.7418300276124,
            "sha256": "a801c56c0f265ea4dad1913c86c55ea9541f3f9ec2db193cbf0ca5420b0121b9",
            "sum": -97346.80173273385,
            "top1": 42
          },
          "gpu_active_fraction": 0.9917355371900827,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 121,
          "gross_j": {
            "core": 9.723059,
            "dram": 4.011342,
            "package-0": 28.605761,
            "psys": 50.163629,
            "uncore": 10.864108
          },
          "net_j": {
            "core": 6.714774087869898,
            "dram": 3.4103233760283933,
            "package-0": 23.344906164557596,
            "psys": 38.00083956741969,
            "uncore": 10.861506953811437
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.224612417,
          "token_latencies_s": [
            0.092988572,
            0.087408365,
            0.087751093,
            0.086844468,
            0.087103207,
            0.08682583,
            0.086671677,
            0.086737236,
            0.086731682,
            0.087889604,
            0.086971591,
            0.086789989,
            0.087064941,
            0.086788371
          ],
          "watts": {
            "core": 7.939703097098353,
            "dram": 3.275601279486292,
            "package-0": 23.359032296991646,
            "psys": 40.96286163983915,
            "uncore": 8.871466473134742
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 15.508871,
            "dram": 5.743332,
            "package-0": 49.502925,
            "psys": 83.542205,
            "uncore": 22.272587
          },
          "net_j": {
            "core": 10.77822681355638,
            "dram": 4.798207006030458,
            "package-0": 41.23002762825894,
            "psys": 64.41574896248571,
            "uncore": 22.268496754431506
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.402521,
            "dram": 0.291564,
            "package-0": 3.743216,
            "psys": 5.430955,
            "uncore": 0.756041
          },
          "logits": {
            "l2": 534.553078595872,
            "sha256": "3565942ee06dde1770dddad80ce01950d2e4f39473f435d47f4fdea8aaa08c44",
            "sum": -125913.45951230347,
            "top1": 546
          },
          "net_j": {
            "core": 2.1729654466803114,
            "dram": 0.24570160105469913,
            "package-0": 3.3417718289991143,
            "psys": 4.502839499088268,
            "uncore": 0.7558425199309685
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.093447459,
          "watts": {
            "core": 25.70985905566464,
            "dram": 3.1200848382618944,
            "package-0": 40.056905132112796,
            "psys": 58.11773865354648,
            "uncore": 8.090546367879302
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 531.7407937505551,
          "decode_tokens_per_s": 11.383173770453885,
          "end_to_end_ms": 1925.750313,
          "handoff_estimate_ms": 6.480631499999986,
          "handoff_plus_first_decode_ms": 93.447459,
          "output_tokens_per_s_e2e": 8.308449902352423,
          "prefill_tokens_per_s": 1657.851428200273,
          "steady_tpot_ms": 86.96682750000001,
          "steady_tpot_p95_ms": 89.9530712,
          "ttft_ms": 608.015883
        },
        "prefill": {
          "gpu_active_fraction": 0.8688524590163934,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 3.908254,
            "dram": 1.439144,
            "package-0": 17.70839,
            "psys": 28.708239,
            "uncore": 10.653354
          },
          "logits": {
            "l2": 878.279080814622,
            "sha256": "952e315a56fbaeb7978e914a4c86da9699720487343b981aaec4300879317855",
            "sum": -272259.80647715216,
            "top1": 198
          },
          "net_j": {
            "core": 2.414650814393594,
            "dram": 1.1407403056713465,
            "package-0": 15.096393539975267,
            "psys": 22.669455582541488,
            "uncore": 10.652062589449928
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.608015883,
          "watts": {
            "core": 6.427881424275228,
            "dram": 2.3669513251843783,
            "package-0": 29.12488060776531,
            "psys": 47.21626490800077,
            "uncore": 17.521506095261003
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 534.7418300276124,
            "sha256": "a801c56c0f265ea4dad1913c86c55ea9541f3f9ec2db193cbf0ca5420b0121b9",
            "sum": -97346.80173273385,
            "top1": 42
          },
          "gpu_active_fraction": 0.9917355371900827,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 121,
          "gross_j": {
            "core": 9.198096,
            "dram": 4.012624,
            "package-0": 28.051319,
            "psys": 49.403011,
            "uncore": 10.863192
          },
          "net_j": {
            "core": 6.190610552482475,
            "dram": 3.411765099304412,
            "package-0": 22.791862259284553,
            "psys": 37.24345388085595,
            "uncore": 10.860591645050611
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.224286971,
          "token_latencies_s": [
            0.092954025,
            0.088337173,
            0.087070661,
            0.086864319,
            0.086993783,
            0.086939872,
            0.08706385,
            0.086786729,
            0.086742508,
            0.086824781,
            0.086553325,
            0.0866952,
            0.087206589,
            0.087199271
          ],
          "watts": {
            "core": 7.5130228597360444,
            "dram": 3.2775191560868127,
            "package-0": 22.912372396716457,
            "psys": 40.35247631496684,
            "uncore": 8.87307653950358
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.555001899600029,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395459,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04405000526458025,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 16.173786,
            "dram": 6.314864,
            "package-0": 46.298099,
            "psys": 81.830112,
            "uncore": 11.602082
          },
          "net_j": {
            "core": 10.151623549275305,
            "dram": 5.11170941351996,
            "package-0": 35.76660900682593,
            "psys": 57.48192160034395,
            "uncore": 11.600541322463972
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 4.724963,
            "dram": 0.507506,
            "package-0": 6.044906,
            "psys": 10.00858,
            "uncore": 0.0
          },
          "logits": {
            "l2": 531.4957347674297,
            "sha256": "ea56709a94624ac035abbd9e592a608814486168ca89335794227ce270636dc5",
            "sum": -124549.59498099051,
            "top1": 546
          },
          "net_j": {
            "core": 3.8510704566173164,
            "dram": 0.3329129314362225,
            "package-0": 4.516652544360215,
            "psys": 6.475347177202146,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 123236,
          "npu_utilization": 0.3464175347135818,
          "seconds": 0.355744117,
          "watts": {
            "core": 13.281914652154317,
            "dram": 1.4266040554087362,
            "package-0": 16.99228662156625,
            "psys": 28.134210860330263,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 417.70319469682715,
          "decode_tokens_per_s": 8.689975332919893,
          "end_to_end_ms": 2451.50148,
          "handoff_estimate_ms": 258.53080500000004,
          "handoff_plus_first_decode_ms": 355.744117,
          "output_tokens_per_s_e2e": 6.526612417137924,
          "prefill_tokens_per_s": 1389.6268198671048,
          "steady_tpot_ms": 97.213312,
          "steady_tpot_p95_ms": 100.73025975,
          "ttft_ms": 725.374601
        },
        "prefill": {
          "gpu_active_fraction": 0.9861111111111112,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 72,
          "gross_j": {
            "core": 4.163258,
            "dram": 1.577083,
            "package-0": 19.398937,
            "psys": 32.816627,
            "uncore": 11.602082
          },
          "logits": {
            "l2": 877.5494660304123,
            "sha256": "5e812874ce95b07d0da2ea5f5b3e9c6c0c235c65f52b97c6d2e966cef93ec2be",
            "sum": -271925.10265737865,
            "top1": 198
          },
          "net_j": {
            "core": 2.3813607747731127,
            "dram": 1.2210816737804135,
            "package-0": 16.28277516446477,
            "psys": 25.612242650118823,
            "uncore": 11.600541322463972
          },
          "npu_busy_delta_us": 14204,
          "npu_utilization": 0.019581606497413053,
          "seconds": 0.725374601,
          "watts": {
            "core": 5.7394593004229,
            "dram": 2.1741635257504695,
            "package-0": 26.743336440587615,
            "psys": 45.24093751664183,
            "uncore": 15.994607453866447
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 528.2898786352387,
            "sha256": "4e0d1a2576ffa7766379d7748d45a9bb599bb9599bebd7b0b95e95ac55cbc585",
            "sum": -93911.72267914936,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 136,
          "gross_j": {
            "core": 7.285565,
            "dram": 4.230275,
            "package-0": 20.854256,
            "psys": 39.004905,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.9191923178848764,
            "dram": 3.5577148083033237,
            "package-0": 14.967181298000943,
            "psys": 25.394331773022984,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1354569,
          "npu_utilization": 0.9884603320776447,
          "seconds": 1.370382762,
          "token_latencies_s": [
            0.106353153,
            0.097702548,
            0.097038749,
            0.096922185,
            0.097211904,
            0.09721472,
            0.097331065,
            0.097288579,
            0.09718452,
            0.097081743,
            0.097358517,
            0.09741837,
            0.097206233,
            0.097040792
          ],
          "watts": {
            "core": 5.316445304206184,
            "dram": 3.0869295187471133,
            "package-0": 15.217832986722874,
            "psys": 28.462781407929008,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.560694000000002,
            "dram": 6.131576,
            "package-0": 43.219311,
            "psys": 75.729237,
            "uncore": 11.600984
          },
          "net_j": {
            "core": 8.106232875038629,
            "dram": 5.041841212376557,
            "package-0": 33.68061070326938,
            "psys": 53.676318453755876,
            "uncore": 11.59944846284036
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.486749,
            "dram": 0.33319,
            "package-0": 3.307914,
            "psys": 5.069506,
            "uncore": 0.0
          },
          "logits": {
            "l2": 531.4957347674297,
            "sha256": "ea56709a94624ac035abbd9e592a608814486168ca89335794227ce270636dc5",
            "sum": -124549.59498099051,
            "top1": 546
          },
          "net_j": {
            "core": 2.165719917559779,
            "dram": 0.26905230592551505,
            "package-0": 2.746501949610992,
            "psys": 3.7715540953414393,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 119789,
          "npu_utilization": 0.9166274711223015,
          "seconds": 0.130684497,
          "watts": {
            "core": 19.028645762014143,
            "dram": 2.5495755628917482,
            "package-0": 25.31221434781204,
            "psys": 38.79194637754162,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 461.17782069054414,
          "decode_tokens_per_s": 10.017047478861741,
          "end_to_end_ms": 2220.401663,
          "handoff_estimate_ms": 33.56858300000001,
          "handoff_plus_first_decode_ms": 130.68449700000002,
          "output_tokens_per_s_e2e": 7.205903448289752,
          "prefill_tokens_per_s": 1394.278745642604,
          "steady_tpot_ms": 97.115914,
          "steady_tpot_p95_ms": 99.6845494,
          "ttft_ms": 722.954433
        },
        "prefill": {
          "gpu_active_fraction": 0.9861111111111112,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 72,
          "gross_j": {
            "core": 3.874623,
            "dram": 1.572689,
            "package-0": 19.126233,
            "psys": 32.137735,
            "uncore": 11.600984
          },
          "logits": {
            "l2": 877.5494660304123,
            "sha256": "5e812874ce95b07d0da2ea5f5b3e9c6c0c235c65f52b97c6d2e966cef93ec2be",
            "sum": -271925.10265737865,
            "top1": 198
          },
          "net_j": {
            "core": 2.0986709659031,
            "dram": 1.2178754504360705,
            "package-0": 16.020468047877543,
            "psys": 24.957387637958277,
            "uncore": 11.59944846284036
          },
          "npu_busy_delta_us": 11317,
          "npu_utilization": 0.015653821988529118,
          "seconds": 0.722954433,
          "watts": {
            "core": 5.359429063767841,
            "dram": 2.1753639347280966,
            "package-0": 26.455654916773987,
            "psys": 44.45333417023255,
            "uncore": 16.046632360853096
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 528.2898786352387,
            "sha256": "4e0d1a2576ffa7766379d7748d45a9bb599bb9599bebd7b0b95e95ac55cbc585",
            "sum": -93911.72267914936,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 135,
          "gross_j": {
            "core": 7.199322,
            "dram": 4.225697,
            "package-0": 20.785164,
            "psys": 38.521996,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.8418419915757505,
            "dram": 3.554913456014971,
            "package-0": 14.913640705780843,
            "psys": 24.947376720456163,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1354409,
          "npu_utilization": 0.9909613185216983,
          "seconds": 1.366762733,
          "token_latencies_s": [
            0.103065033,
            0.097864289,
            0.097126703,
            0.097512483,
            0.097448986,
            0.097454217,
            0.097148124,
            0.097019804,
            0.097105125,
            0.096955413,
            0.097014231,
            0.096994465,
            0.096999023,
            0.097008555
          ],
          "watts": {
            "core": 5.26742632512208,
            "dram": 3.0917560875578833,
            "package-0": 15.207587606941285,
            "psys": 28.184845159953596,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.908657,
            "dram": 6.135421,
            "package-0": 43.591929,
            "psys": 76.838182,
            "uncore": 11.602021
          },
          "net_j": {
            "core": 8.451572439402389,
            "dram": 5.045162081942696,
            "package-0": 34.04864086852117,
            "psys": 54.774656647497316,
            "uncore": 11.600481525806272
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.575311,
            "dram": 0.333373,
            "package-0": 3.406729,
            "psys": 5.199755,
            "uncore": 0.0
          },
          "logits": {
            "l2": 531.4957347674297,
            "sha256": "ea56709a94624ac035abbd9e592a608814486168ca89335794227ce270636dc5",
            "sum": -124549.59498099051,
            "top1": 546
          },
          "net_j": {
            "core": 2.2572617795109746,
            "dram": 0.2698306459886558,
            "package-0": 2.850528098691881,
            "psys": 3.913850967969597,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 119722,
          "npu_utilization": 0.9246980347907574,
          "seconds": 0.129471455,
          "watts": {
            "core": 19.89095588676284,
            "dram": 2.574876446704024,
            "package-0": 26.312587589287535,
            "psys": 40.16140082769595,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 460.95611433507025,
          "decode_tokens_per_s": 10.022305897904792,
          "end_to_end_ms": 2221.469611,
          "handoff_estimate_ms": 32.22111100000001,
          "handoff_plus_first_decode_ms": 129.47145500000002,
          "output_tokens_per_s_e2e": 7.202439286485473,
          "prefill_tokens_per_s": 1390.7130327687628,
          "steady_tpot_ms": 97.250344,
          "steady_tpot_p95_ms": 99.3982875,
          "ttft_ms": 724.808049
        },
        "prefill": {
          "gpu_active_fraction": 0.9861111111111112,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 72,
          "gross_j": {
            "core": 4.160329,
            "dram": 1.5733,
            "package-0": 19.428844,
            "psys": 32.394814,
            "uncore": 11.602021
          },
          "logits": {
            "l2": 877.5494660304123,
            "sha256": "5e812874ce95b07d0da2ea5f5b3e9c6c0c235c65f52b97c6d2e966cef93ec2be",
            "sum": -271925.10265737865,
            "top1": 198
          },
          "net_j": {
            "core": 2.379823521100259,
            "dram": 1.2175767277170888,
            "package-0": 16.315116034795786,
            "psys": 25.196056616318863,
            "uncore": 11.600481525806272
          },
          "npu_busy_delta_us": 11733,
          "npu_utilization": 0.01618773414035307,
          "seconds": 0.724808049,
          "watts": {
            "core": 5.739904524708169,
            "dram": 2.170643665134022,
            "package-0": 26.805502542094427,
            "psys": 44.69433534119045,
            "uncore": 16.007025606306424
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 528.2898786352387,
            "sha256": "4e0d1a2576ffa7766379d7748d45a9bb599bb9599bebd7b0b95e95ac55cbc585",
            "sum": -93911.72267914936,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 136,
          "gross_j": {
            "core": 7.173017,
            "dram": 4.228748,
            "package-0": 20.756356,
            "psys": 39.243613,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.8144871387911548,
            "dram": 3.5577547082369514,
            "package-0": 14.8829967350335,
            "psys": 25.664749063208856,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1353782,
          "npu_utilization": 0.990192946151855,
          "seconds": 1.367190107,
          "token_latencies_s": [
            0.102315247,
            0.097827617,
            0.097505718,
            0.09739017,
            0.09711434,
            0.097241242,
            0.097330274,
            0.097231508,
            0.097003388,
            0.097155065,
            0.09736003,
            0.097210354,
            0.097219653,
            0.097259446
          ],
          "watts": {
            "core": 5.246539572861318,
            "dram": 3.0930212106925374,
            "package-0": 15.181762868036902,
            "psys": 28.703845060809822,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 33.750633016228676,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395594,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU+GPU",
      "key": "npu_gpu_combined",
      "label": "NPU + iGPU (subgraph combination)",
      "mode": "combined",
      "prefill_device": "NPU+GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.043780725449323654,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 107.71627699999999,
            "dram": 19.438302999999998,
            "package-0": 234.553965,
            "psys": 404.213076,
            "uncore": 65.215653
          },
          "net_j": {
            "core": 77.11498131454753,
            "dram": 13.32453746188281,
            "package-0": 181.03876312252137,
            "psys": 280.48905162800594,
            "uncore": 65.1891942749016
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5337837837837838,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 148,
          "gross_j": {
            "core": 17.623185,
            "dram": 1.700923,
            "package-0": 26.469476,
            "psys": 43.057812,
            "uncore": 4.252552
          },
          "logits": {
            "l2": 504.4882724745259,
            "sha256": "c2c1db075c1b7545142fc4eaec7eab893a508b06b424cdfbd622a43353872895",
            "sum": -101006.92088515963,
            "top1": 287
          },
          "net_j": {
            "core": 13.932972281353123,
            "dram": 0.9636635198953609,
            "package-0": 20.01607348269718,
            "psys": 28.137921906694036,
            "uncore": 4.249361340203077
          },
          "npu_busy_delta_us": 14192,
          "npu_utilization": 0.0094474044455185,
          "seconds": 1.502211542,
          "watts": {
            "core": 11.731493539543047,
            "dram": 1.1322792778808246,
            "package-0": 17.62033858743977,
            "psys": 28.66294845709553,
            "uncore": 2.830860954734962
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 82.20163356830707,
          "decode_tokens_per_s": 1.2894316514781647,
          "end_to_end_ms": 12457.173362000001,
          "handoff_estimate_ms": 780.9984665,
          "handoff_plus_first_decode_ms": 1502.211542,
          "output_tokens_per_s_e2e": 1.284400524504798,
          "prefill_tokens_per_s": 1223.0915298083162,
          "steady_tpot_ms": 721.2130755000001,
          "steady_tpot_p95_ms": 738.7687036,
          "ttft_ms": 824.141101
        },
        "prefill": {
          "gpu_active_fraction": 0.06097560975609756,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 82,
          "gross_j": {
            "core": 3.074211,
            "dram": 1.591731,
            "package-0": 18.341872,
            "psys": 28.927233,
            "uncore": 0.336242
          },
          "logits": {
            "l2": 868.0879947489402,
            "sha256": "8f622f28568639fbc823d20d44054c57cca137aa922702e247efb4f1b0ff4d7a",
            "sum": -268991.91000521276,
            "top1": 198
          },
          "net_j": {
            "core": 1.0496918907800015,
            "dram": 1.1872567813775174,
            "package-0": 14.80141574127281,
            "psys": 20.741904699349675,
            "uncore": 0.33449154488470256
          },
          "npu_busy_delta_us": 736047,
          "npu_utilization": 0.893107987342085,
          "seconds": 0.824141101,
          "watts": {
            "core": 3.730199836253525,
            "dram": 1.9313816506282944,
            "package-0": 22.25574234526619,
            "psys": 35.099854824495644,
            "uncore": 0.40799081564068235
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1138.6761954498609,
            "sha256": "8f8ad56dcbeddedefeae83c67e079a5e7b4a1b66f1c7242fd23b39a60742d69c",
            "sum": -356217.6366803106,
            "top1": 90
          },
          "gpu_active_fraction": 0.993993993993994,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 999,
          "gross_j": {
            "core": 87.018881,
            "dram": 16.145649,
            "package-0": 189.742617,
            "psys": 332.228031,
            "uncore": 60.626859
          },
          "net_j": {
            "core": 62.132317142414415,
            "dram": 11.173617160609933,
            "package-0": 146.22127389855137,
            "psys": 231.60922502196223,
            "uncore": 60.605341389813816
          },
          "npu_busy_delta_us": 201749,
          "npu_utilization": 0.019914378666441784,
          "seconds": 10.130820719,
          "token_latencies_s": [
            0.73201316,
            0.713476546,
            0.712860235,
            0.715058908,
            0.721500303,
            0.720925848,
            0.724050688,
            0.713012982,
            0.719032783,
            0.720298478,
            0.725911012,
            0.735033955,
            0.738717584,
            0.73886364
          ],
          "watts": {
            "core": 8.589519389756756,
            "dram": 1.5937157953767158,
            "package-0": 18.729244378408982,
            "psys": 32.79379235059583,
            "uncore": 5.984397580572761
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 97.409784,
            "dram": 20.292367,
            "package-0": 227.927028,
            "psys": 400.283826,
            "uncore": 67.829844
          },
          "net_j": {
            "core": 67.9407518651261,
            "dram": 14.404813905524023,
            "package-0": 176.39191589095978,
            "psys": 281.1376536433995,
            "uncore": 67.80436426122726
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9868421052631579,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 76,
          "gross_j": {
            "core": 8.620278,
            "dram": 1.259274,
            "package-0": 16.349201,
            "psys": 27.453604,
            "uncore": 4.464588
          },
          "logits": {
            "l2": 1153.8101991094447,
            "sha256": "122341610e500cb6d0a918eba9b1179dddb1ce717cb641f1095dfb29faf2c7e7",
            "sum": -361862.7407342456,
            "top1": 90
          },
          "net_j": {
            "core": 6.741920964479739,
            "dram": 0.8840011805061825,
            "package-0": 13.064351332616944,
            "psys": 19.85922322505474,
            "uncore": 4.462963920648905
          },
          "npu_busy_delta_us": 11985,
          "npu_utilization": 0.01567401288185979,
          "seconds": 0.764641454,
          "watts": {
            "core": 11.27362106109408,
            "dram": 1.6468816769120653,
            "package-0": 21.381525830798076,
            "psys": 35.903891760490396,
            "uncore": 5.838799317830341
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 85.3599970008565,
          "decode_tokens_per_s": 1.3420812457946254,
          "end_to_end_ms": 11996.251593,
          "handoff_estimate_ms": 19.12891849999998,
          "handoff_plus_first_decode_ms": 764.6414540000001,
          "output_tokens_per_s_e2e": 1.3337499531383827,
          "prefill_tokens_per_s": 1229.8967474227486,
          "steady_tpot_ms": 745.5125355000001,
          "steady_tpot_p95_ms": 751.2121039499999,
          "ttft_ms": 819.580995
        },
        "prefill": {
          "gpu_active_fraction": 0.04938271604938271,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 81,
          "gross_j": {
            "core": 3.345633,
            "dram": 1.58929,
            "package-0": 18.775525,
            "psys": 28.453907,
            "uncore": 0.449584
          },
          "logits": {
            "l2": 868.0879947489402,
            "sha256": "8f622f28568639fbc823d20d44054c57cca137aa922702e247efb4f1b0ff4d7a",
            "sum": -268991.91000521276,
            "top1": 198
          },
          "net_j": {
            "core": 1.3323158824363652,
            "dram": 1.187053802644686,
            "package-0": 15.254658658349594,
            "psys": 20.313869444648596,
            "uncore": 0.44784323043595625
          },
          "npu_busy_delta_us": 740369,
          "npu_utilization": 0.903350620032374,
          "seconds": 0.819580995,
          "watts": {
            "core": 4.082126135684734,
            "dram": 1.9391494064598216,
            "package-0": 22.90868762763343,
            "psys": 34.71762665750931,
            "uncore": 0.5485534715211399
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1303.0924465167705,
            "sha256": "55d57314a17ef9ab9c3edbfcf30a0f6d7df14c83b86296dc6078fa81b9923749",
            "sum": -420117.63300922886,
            "top1": 90
          },
          "gpu_active_fraction": 0.9931573802541545,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1023,
          "gross_j": {
            "core": 85.443873,
            "dram": 17.443803,
            "package-0": 192.802302,
            "psys": 344.376315,
            "uncore": 62.915672
          },
          "net_j": {
            "core": 59.86651501821,
            "dram": 12.333758922373153,
            "package-0": 148.07290589999323,
            "psys": 240.96456097369617,
            "uncore": 62.893557110142396
          },
          "npu_busy_delta_us": 201261,
          "npu_utilization": 0.01932966160740896,
          "seconds": 10.412029144,
          "token_latencies_s": [
            0.753083725,
            0.733468279,
            0.734218112,
            0.733895279,
            0.733983938,
            0.744130347,
            0.745302433,
            0.742140772,
            0.745722638,
            0.748726863,
            0.748456491,
            0.749087426,
            0.749552264,
            0.750204308
          ],
          "watts": {
            "core": 8.206265255148425,
            "dram": 1.6753509578920172,
            "package-0": 18.517264918635345,
            "psys": 33.07485123574103,
            "uncore": 6.042594688303919
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 99.978931,
            "dram": 20.561775,
            "package-0": 231.591082,
            "psys": 414.558204,
            "uncore": 67.96882000000001
          },
          "net_j": {
            "core": 70.16604767755804,
            "dram": 14.605524633277684,
            "package-0": 179.4546467996874,
            "psys": 294.02180774084644,
            "uncore": 67.94304295799735
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.987012987012987,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 77,
          "gross_j": {
            "core": 8.253824,
            "dram": 1.301388,
            "package-0": 16.139302,
            "psys": 27.746633,
            "uncore": 4.528431
          },
          "logits": {
            "l2": 1309.6928914418934,
            "sha256": "57562c01e4413e6fbaea8390321d7b602247a4cea9321eea3fedec8e1740f1c0",
            "sum": -422736.6517739836,
            "top1": 90
          },
          "net_j": {
            "core": 6.329860808175593,
            "dram": 0.9170036268777255,
            "package-0": 12.774696799172899,
            "psys": 19.967862085104905,
            "uncore": 4.526767488307057
          },
          "npu_busy_delta_us": 11826,
          "npu_utilization": 0.015099460224940206,
          "seconds": 0.783206805,
          "watts": {
            "core": 10.538498832374165,
            "dram": 1.6616147761892852,
            "package-0": 20.606692762328592,
            "psys": 35.42695597492926,
            "uncore": 5.781909670715898
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 84.37548517011194,
          "decode_tokens_per_s": 1.325303811963323,
          "end_to_end_ms": 12136.226511,
          "handoff_estimate_ms": 31.28212149999998,
          "handoff_plus_first_decode_ms": 783.2068049999999,
          "output_tokens_per_s_e2e": 1.3183669557829991,
          "prefill_tokens_per_s": 1232.1729781928193,
          "steady_tpot_ms": 751.9246835,
          "steady_tpot_p95_ms": 757.3541970499999,
          "ttft_ms": 818.066958
        },
        "prefill": {
          "gpu_active_fraction": 0.06172839506172839,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 81,
          "gross_j": {
            "core": 3.045341,
            "dram": 1.59222,
            "package-0": 18.398024,
            "psys": 28.837207,
            "uncore": 0.341307
          },
          "logits": {
            "l2": 868.0879947489402,
            "sha256": "8f622f28568639fbc823d20d44054c57cca137aa922702e247efb4f1b0ff4d7a",
            "sum": -268991.91000521276,
            "top1": 198
          },
          "net_j": {
            "core": 1.0357431446292682,
            "dram": 1.1907268658553394,
            "package-0": 14.883661862260904,
            "psys": 20.712206784600813,
            "uncore": 0.33956944621273066
          },
          "npu_busy_delta_us": 735297,
          "npu_utilization": 0.8988225142323863,
          "seconds": 0.818066958,
          "watts": {
            "core": 3.722606041252677,
            "dram": 1.9463199001370741,
            "package-0": 22.48963097712596,
            "psys": 35.25042384122303,
            "uncore": 0.41721157010719906
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1378.3485702204835,
            "sha256": "daee1e6fdbbca988a1804f628833400f18d9971e52c6f7bf77318beff12d729e",
            "sum": -450508.6694248039,
            "top1": 90
          },
          "gpu_active_fraction": 0.9951969260326609,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1041,
          "gross_j": {
            "core": 88.679766,
            "dram": 17.668167,
            "package-0": 197.053756,
            "psys": 357.974364,
            "uncore": 63.099082
          },
          "net_j": {
            "core": 62.80044372475318,
            "dram": 12.497794140544618,
            "package-0": 151.7962881382536,
            "psys": 253.3417388711407,
            "uncore": 63.07670602347756
          },
          "npu_busy_delta_us": 200000,
          "npu_utilization": 0.018984423070902607,
          "seconds": 10.534952748,
          "token_latencies_s": [
            0.76036756,
            0.74992104,
            0.747467182,
            0.75232368,
            0.749285046,
            0.749908574,
            0.755383249,
            0.755731617,
            0.751335529,
            0.754336254,
            0.754344248,
            0.752925592,
            0.75004468,
            0.751525687
          ],
          "watts": {
            "core": 8.417670977863223,
            "dram": 1.6770997860768004,
            "package-0": 18.704759358072064,
            "psys": 33.97968387356643,
            "uncore": 5.989498340367876
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.377430791035295,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395298,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.047323985025286674,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 16.195636999999998,
            "dram": 5.704088,
            "package-0": 50.023065,
            "psys": 83.195099,
            "uncore": 22.174504
          },
          "net_j": {
            "core": 11.512037060894261,
            "dram": 4.768361872702556,
            "package-0": 41.832438079135876,
            "psys": 64.2588474448064,
            "uncore": 22.170454430186567
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.60357,
            "dram": 0.296753,
            "package-0": 3.974172,
            "psys": 5.712083,
            "uncore": 0.770018
          },
          "logits": {
            "l2": 534.553078595872,
            "sha256": "3565942ee06dde1770dddad80ce01950d2e4f39473f435d47f4fdea8aaa08c44",
            "sum": -125913.45951230347,
            "top1": 546
          },
          "net_j": {
            "core": 2.3612137247769156,
            "dram": 0.24833317298889318,
            "package-0": 3.550342070439115,
            "psys": 4.73221293140479,
            "uncore": 0.7698084520734051
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.098658376,
          "watts": {
            "core": 26.389751236124138,
            "dram": 3.0078845003489616,
            "package-0": 40.2821550600022,
            "psys": 57.89759807114603,
            "uncore": 7.804892308383425
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 537.0818445973571,
          "decode_tokens_per_s": 11.343513880930132,
          "end_to_end_ms": 1906.5995440000002,
          "handoff_estimate_ms": 11.587197500000007,
          "handoff_plus_first_decode_ms": 98.658376,
          "output_tokens_per_s_e2e": 8.391903821833704,
          "prefill_tokens_per_s": 1725.2652967660842,
          "steady_tpot_ms": 87.0711785,
          "steady_tpot_p95_ms": 89.24730315000001,
          "ttft_ms": 584.25797
        },
        "prefill": {
          "gpu_active_fraction": 0.9137931034482759,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 4.215443,
            "dram": 1.402157,
            "package-0": 17.850601,
            "psys": 28.261158,
            "uncore": 10.581516
          },
          "logits": {
            "l2": 878.279080814622,
            "sha256": "952e315a56fbaeb7978e914a4c86da9699720487343b981aaec4300879317855",
            "sum": -272259.80647715216,
            "top1": 198
          },
          "net_j": {
            "core": 2.7802016031762724,
            "dram": 1.115413279048619,
            "package-0": 15.340666980100002,
            "psys": 22.458336989230496,
            "uncore": 10.580275050663777
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.58425797,
          "watts": {
            "core": 7.215037220630468,
            "dram": 2.3998936634103596,
            "package-0": 30.55260161876782,
            "psys": 48.37102692839603,
            "uncore": 18.111034069419713
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 534.7418300276124,
            "sha256": "a801c56c0f265ea4dad1913c86c55ea9541f3f9ec2db193cbf0ca5420b0121b9",
            "sum": -97346.80173273385,
            "top1": 42
          },
          "gpu_active_fraction": 0.9918032786885246,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 122,
          "gross_j": {
            "core": 9.376624,
            "dram": 4.005178,
            "package-0": 28.198292,
            "psys": 49.221858,
            "uncore": 10.82297
          },
          "net_j": {
            "core": 6.370621732941074,
            "dram": 3.4046154206650434,
            "package-0": 22.941429028596758,
            "psys": 37.068297524171115,
            "uncore": 10.820370927449385
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.223683198,
          "token_latencies_s": [
            0.091057793,
            0.088272424,
            0.087096806,
            0.087255071,
            0.087049273,
            0.086903285,
            0.087185495,
            0.087029006,
            0.086999301,
            0.086927765,
            0.087093084,
            0.086924588,
            0.087160468,
            0.086700812
          ],
          "watts": {
            "core": 7.662623802733622,
            "dram": 3.2730513964285057,
            "package-0": 23.04378457274527,
            "psys": 40.2243473477847,
            "uncore": 8.844584952779583
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.605548,
            "dram": 5.750108,
            "package-0": 50.596977,
            "psys": 84.290678,
            "uncore": 22.246525
          },
          "net_j": {
            "core": 11.852039633153733,
            "dram": 4.800415021732021,
            "package-0": 42.28409500660801,
            "psys": 65.07177985014967,
            "uncore": 22.24241498542989
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.459283,
            "dram": 0.292297,
            "package-0": 3.797171,
            "psys": 5.481676,
            "uncore": 0.753355
          },
          "logits": {
            "l2": 534.553078595872,
            "sha256": "3565942ee06dde1770dddad80ce01950d2e4f39473f435d47f4fdea8aaa08c44",
            "sum": -125913.45951230347,
            "top1": 546
          },
          "net_j": {
            "core": 2.229860464782566,
            "dram": 0.24646117644869125,
            "package-0": 3.395959449560959,
            "psys": 4.554098304254901,
            "uncore": 0.7531566349420922
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.09339331,
          "watts": {
            "core": 26.332539236482788,
            "dram": 3.1297423766220516,
            "package-0": 40.65784797647712,
            "psys": 58.694525335915394,
            "uncore": 8.066477138458847
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 529.183142328175,
          "decode_tokens_per_s": 11.418517217540597,
          "end_to_end_ms": 1935.057862,
          "handoff_estimate_ms": 6.424753999999991,
          "handoff_plus_first_decode_ms": 93.39331,
          "output_tokens_per_s_e2e": 8.268486598877734,
          "prefill_tokens_per_s": 1622.1378508130183,
          "steady_tpot_ms": 86.968556,
          "steady_tpot_p95_ms": 88.41024125,
          "ttft_ms": 621.4021819999999
        },
        "prefill": {
          "gpu_active_fraction": 0.8548387096774194,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 4.44994,
            "dram": 1.450558,
            "package-0": 18.286452,
            "psys": 29.228624,
            "uncore": 10.660495
          },
          "logits": {
            "l2": 878.279080814622,
            "sha256": "952e315a56fbaeb7978e914a4c86da9699720487343b981aaec4300879317855",
            "sum": -272259.80647715216,
            "top1": 198
          },
          "net_j": {
            "core": 2.923453102982588,
            "dram": 1.1455845413170362,
            "package-0": 15.61694887513567,
            "psys": 23.056888528684798,
            "uncore": 10.659175157285178
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.621402182,
          "watts": {
            "core": 7.1611270911179385,
            "dram": 2.334330393452014,
            "package-0": 29.427724153051013,
            "psys": 47.0365647991883,
            "uncore": 17.15554806339576
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 534.7418300276124,
            "sha256": "a801c56c0f265ea4dad1913c86c55ea9541f3f9ec2db193cbf0ca5420b0121b9",
            "sum": -97346.80173273385,
            "top1": 42
          },
          "gpu_active_fraction": 0.9917355371900827,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 121,
          "gross_j": {
            "core": 9.696325,
            "dram": 4.007253,
            "package-0": 28.513354,
            "psys": 49.580378,
            "uncore": 10.832675
          },
          "net_j": {
            "core": 6.698726065388579,
            "dram": 3.408369303966293,
            "package-0": 23.27118668191138,
            "psys": 37.46079301720997,
            "uncore": 10.83008319320262
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.22026237,
          "token_latencies_s": [
            0.08960967,
            0.087240858,
            0.087345714,
            0.087764395,
            0.086776331,
            0.087085376,
            0.086990121,
            0.086680053,
            0.086946991,
            0.086553458,
            0.087400396,
            0.086636014,
            0.086724405,
            0.086483082
          ],
          "watts": {
            "core": 7.946098509945857,
            "dram": 3.283927373749959,
            "package-0": 23.366576484694846,
            "psys": 40.630916120112765,
            "uncore": 8.877332667400045
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.382099999999998,
            "dram": 5.755051,
            "package-0": 50.564812,
            "psys": 84.393583,
            "uncore": 22.405399
          },
          "net_j": {
            "core": 11.63086274731996,
            "dram": 4.805811762634541,
            "package-0": 42.255901705530235,
            "psys": 65.18386718632726,
            "uncore": 22.401290949097902
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.453241,
            "dram": 0.293945,
            "package-0": 3.746633,
            "psys": 5.442491,
            "uncore": 0.697752
          },
          "logits": {
            "l2": 534.553078595872,
            "sha256": "3565942ee06dde1770dddad80ce01950d2e4f39473f435d47f4fdea8aaa08c44",
            "sum": -125913.45951230347,
            "top1": 546
          },
          "net_j": {
            "core": 2.218598995990846,
            "dram": 0.24706639026231209,
            "package-0": 3.3362937012148364,
            "psys": 4.493810482641285,
            "uncore": 0.6975491220452744
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.095518051,
          "watts": {
            "core": 25.683532843441284,
            "dram": 3.07737644269982,
            "package-0": 39.224345144982074,
            "psys": 56.97866469239411,
            "uncore": 7.304922919752624
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 529.4360944051087,
          "decode_tokens_per_s": 11.391117970778227,
          "end_to_end_ms": 1934.133337,
          "handoff_estimate_ms": 8.411922000000002,
          "handoff_plus_first_decode_ms": 95.518051,
          "output_tokens_per_s_e2e": 8.272438975079824,
          "prefill_tokens_per_s": 1632.8701980500973,
          "steady_tpot_ms": 87.10612900000001,
          "steady_tpot_p95_ms": 87.98141995,
          "ttft_ms": 617.317899
        },
        "prefill": {
          "gpu_active_fraction": 0.8524590163934426,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 4.498158,
            "dram": 1.449581,
            "package-0": 18.488722,
            "psys": 29.511216,
            "uncore": 10.805026
          },
          "logits": {
            "l2": 878.279080814622,
            "sha256": "952e315a56fbaeb7978e914a4c86da9699720487343b981aaec4300879317855",
            "sum": -272259.80647715216,
            "top1": 198
          },
          "net_j": {
            "core": 2.9817042259161846,
            "dram": 1.146612036873748,
            "package-0": 15.836764687730401,
            "psys": 23.380045423685424,
            "uncore": 10.803714832200265
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.617317899,
          "watts": {
            "core": 7.286615222540307,
            "dram": 2.3481920779361687,
            "package-0": 29.950082493882135,
            "psys": 47.805540788312705,
            "uncore": 17.503179508488543
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 534.7418300276124,
            "sha256": "a801c56c0f265ea4dad1913c86c55ea9541f3f9ec2db193cbf0ca5420b0121b9",
            "sum": -97346.80173273385,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 121,
          "gross_j": {
            "core": 9.430701,
            "dram": 4.011525,
            "package-0": 28.329457,
            "psys": 49.439876,
            "uncore": 10.902621
          },
          "net_j": {
            "core": 6.430559525412929,
            "dram": 3.4121333354984804,
            "package-0": 23.082843316584995,
            "psys": 37.31001128000056,
            "uncore": 10.900026994852363
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.221297387,
          "token_latencies_s": [
            0.088140902,
            0.087087482,
            0.08706234,
            0.087106793,
            0.087015262,
            0.087121043,
            0.087105465,
            0.087113481,
            0.086876182,
            0.08772084,
            0.086807005,
            0.087296786,
            0.087895545,
            0.086920479
          ],
          "watts": {
            "core": 7.721871102308352,
            "dram": 3.284642252329653,
            "package-0": 23.196198814105873,
            "psys": 40.481439267993785,
            "uncore": 8.9270812465924
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.5578020960092545,
    "compile_target": "AUTO:GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 403449,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04383872449398041,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 20.020274,
            "dram": 7.308209,
            "package-0": 70.552737,
            "psys": 117.574224,
            "uncore": 35.377962
          },
          "net_j": {
            "core": 10.146228027033347,
            "dram": 6.015469369054168,
            "package-0": 55.560811919046664,
            "psys": 90.98584646674911,
            "uncore": 35.37284906262151
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.416742,
            "dram": 0.305541,
            "package-0": 3.810903,
            "psys": 5.626634,
            "uncore": 0.768004
          },
          "logits": {
            "l2": 529.3319535900215,
            "sha256": "f984afc7983df3c72b7c030f6eb4d656d8481c00cf53a50b1f34deda53645068",
            "sum": -138429.60764685227,
            "top1": 319
          },
          "net_j": {
            "core": 2.0202993930298403,
            "dram": 0.25363754818005385,
            "package-0": 3.208977722472782,
            "psys": 4.559111554464755,
            "uncore": 0.7677987157397128
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.105108209,
          "watts": {
            "core": 22.992894874652468,
            "dram": 2.906918526220916,
            "package-0": 36.25694925502917,
            "psys": 53.531822619106755,
            "uncore": 7.30679370628416
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 782.3093029721824,
          "decode_tokens_per_s": 11.196968461695255,
          "end_to_end_ms": 2617.8903819999996,
          "handoff_estimate_ms": 17.27638549999999,
          "handoff_plus_first_decode_ms": 105.10820899999999,
          "output_tokens_per_s_e2e": 6.111791429470175,
          "prefill_tokens_per_s": 1589.6832548369725,
          "steady_tpot_ms": 87.8318235,
          "steady_tpot_p95_ms": 89.73360699999999,
          "ttft_ms": 1278.2420610000001
        },
        "prefill": {
          "gpu_active_fraction": 0.905511811023622,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 127,
          "gross_j": {
            "core": 9.087379,
            "dram": 2.925225,
            "package-0": 39.002159,
            "psys": 62.381066,
            "uncore": 23.47339
          },
          "logits": {
            "l2": 813.1669789722381,
            "sha256": "d50d0ae7fdda5cdbe691764d1ca1d7d8554922da6b22bef84ee4b7e361ab62dd",
            "sum": -253097.8784024875,
            "top1": 319
          },
          "net_j": {
            "core": 4.266160754318068,
            "dram": 2.2940166875518053,
            "package-0": 31.68202468666213,
            "psys": 49.39871092110475,
            "uncore": 23.470893496839523
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.278242061,
          "watts": {
            "core": 7.109278654850961,
            "dram": 2.288474999572088,
            "package-0": 30.512342059443466,
            "psys": 48.80223230269685,
            "uncore": 18.36380660298112
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 553.0736012792826,
            "sha256": "7bd1634f8d377902a10467f3d09341ea554283d2d40c1594e26340bce4ce43e1",
            "sum": -121082.53280104871,
            "top1": 42
          },
          "gpu_active_fraction": 0.9917355371900827,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 121,
          "gross_j": {
            "core": 8.516153,
            "dram": 4.077443,
            "package-0": 27.739675,
            "psys": 49.566524,
            "uncore": 11.136568
          },
          "net_j": {
            "core": 3.859767879685439,
            "dram": 3.4678151333223086,
            "package-0": 20.66980950991175,
            "psys": 37.0280239911796,
            "uncore": 11.134156850042277
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.234540112,
          "token_latencies_s": [
            0.09163682,
            0.08838888,
            0.087822854,
            0.087666229,
            0.087725201,
            0.087814205,
            0.08784324,
            0.0887088,
            0.087757841,
            0.087646695,
            0.087840793,
            0.087761656,
            0.08791166,
            0.087982589
          ],
          "watts": {
            "core": 6.8982392043977585,
            "dram": 3.302803173721422,
            "package-0": 22.469642525475106,
            "psys": 40.14978818282415,
            "uncore": 9.020823132233716
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 20.793465,
            "dram": 7.380598,
            "package-0": 71.720886,
            "psys": 119.79138,
            "uncore": 35.54245
          },
          "net_j": {
            "core": 10.707521335106975,
            "dram": 6.060116089464915,
            "package-0": 56.407233196602945,
            "psys": 92.63241409327102,
            "uncore": 35.53722733864085
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.609307,
            "dram": 0.300232,
            "package-0": 3.907461,
            "psys": 5.690049,
            "uncore": 0.69055
          },
          "logits": {
            "l2": 529.3319535900215,
            "sha256": "f984afc7983df3c72b7c030f6eb4d656d8481c00cf53a50b1f34deda53645068",
            "sum": -138429.60764685227,
            "top1": 319
          },
          "net_j": {
            "core": 2.231430726554124,
            "dram": 0.2507593080948762,
            "package-0": 3.3337252895444887,
            "psys": 4.672521124725359,
            "uncore": 0.6903543296812438
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.100185746,
          "watts": {
            "core": 26.04469302449472,
            "dram": 2.9967536499653353,
            "package-0": 39.00216503852754,
            "psys": 56.79499556753313,
            "uncore": 6.892697090861608
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 765.8736038269113,
          "decode_tokens_per_s": 11.225396339072269,
          "end_to_end_ms": 2674.0704859999996,
          "handoff_estimate_ms": 12.505427500000014,
          "handoff_plus_first_decode_ms": 100.18574600000001,
          "output_tokens_per_s_e2e": 5.983387529897745,
          "prefill_tokens_per_s": 1518.8948743346662,
          "steady_tpot_ms": 87.6803185,
          "steady_tpot_p95_ms": 90.99372635,
          "ttft_ms": 1337.814772
        },
        "prefill": {
          "gpu_active_fraction": 0.8796992481203008,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 133,
          "gross_j": {
            "core": 8.266031,
            "dram": 3.011101,
            "package-0": 38.60946,
            "psys": 62.960532,
            "uncore": 23.686951
          },
          "logits": {
            "l2": 813.1669789722381,
            "sha256": "d50d0ae7fdda5cdbe691764d1ca1d7d8554922da6b22bef84ee4b7e361ab62dd",
            "sum": -253097.8784024875,
            "top1": 319
          },
          "net_j": {
            "core": 3.2201189737103135,
            "dram": 2.350475105760719,
            "package-0": 30.948169454024544,
            "psys": 49.37313182768218,
            "uncore": 23.684338146841487
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.337814772,
          "watts": {
            "core": 6.178755963086346,
            "dram": 2.250760765257868,
            "package-0": 28.860093944305767,
            "psys": 47.06221916347624,
            "uncore": 17.70570298352185
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 553.0736012792826,
            "sha256": "7bd1634f8d377902a10467f3d09341ea554283d2d40c1594e26340bce4ce43e1",
            "sum": -121082.53280104871,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 122,
          "gross_j": {
            "core": 9.918127,
            "dram": 4.069265,
            "package-0": 29.203965,
            "psys": 51.140799,
            "uncore": 11.164949
          },
          "net_j": {
            "core": 5.255971634842537,
            "dram": 3.45888167560932,
            "package-0": 22.125338453033915,
            "psys": 38.58676114086349,
            "uncore": 11.162534862118115
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.236069968,
          "token_latencies_s": [
            0.096507482,
            0.088024781,
            0.087936399,
            0.087922903,
            0.087716069,
            0.087548329,
            0.087646186,
            0.087683396,
            0.087677241,
            0.087499296,
            0.087494305,
            0.08732335,
            0.087683446,
            0.087384418
          ],
          "watts": {
            "core": 8.023920374060896,
            "dram": 3.2920992381881087,
            "package-0": 23.626465941287233,
            "psys": 41.37370887082341,
            "uncore": 9.032618936665243
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 19.931467,
            "dram": 7.370525,
            "package-0": 70.980775,
            "psys": 119.33660900000001,
            "uncore": 35.634918
          },
          "net_j": {
            "core": 9.861668672205102,
            "dram": 6.052156885286696,
            "package-0": 55.6916359253059,
            "psys": 92.22111851553885,
            "uncore": 35.629703698952085
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.385187,
            "dram": 0.301207,
            "package-0": 3.681936,
            "psys": 5.470994,
            "uncore": 0.685301
          },
          "logits": {
            "l2": 529.3319535900215,
            "sha256": "f984afc7983df3c72b7c030f6eb4d656d8481c00cf53a50b1f34deda53645068",
            "sum": -138429.60764685227,
            "top1": 319
          },
          "net_j": {
            "core": 2.0087389814911294,
            "dram": 0.25192129950297937,
            "package-0": 3.1103688323151726,
            "psys": 4.457312064027524,
            "uncore": 0.6851060692542692
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.099807075,
          "watts": {
            "core": 23.897975168594012,
            "dram": 3.017892268659311,
            "package-0": 36.8905310570418,
            "psys": 54.815693176059916,
            "uncore": 6.866256725788228
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 767.1015616375668,
          "decode_tokens_per_s": 11.229115187630274,
          "end_to_end_ms": 2669.789898,
          "handoff_estimate_ms": 12.054259999999983,
          "handoff_plus_first_decode_ms": 99.807075,
          "output_tokens_per_s_e2e": 5.99298095029349,
          "prefill_tokens_per_s": 1523.2649591568136,
          "steady_tpot_ms": 87.75281500000001,
          "steady_tpot_p95_ms": 90.78791505,
          "ttft_ms": 1333.976724
        },
        "prefill": {
          "gpu_active_fraction": 0.8863636363636364,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 132,
          "gross_j": {
            "core": 8.293618,
            "dram": 3.000542,
            "package-0": 38.77578,
            "psys": 63.534201,
            "uncore": 23.806884
          },
          "logits": {
            "l2": 813.1669789722381,
            "sha256": "d50d0ae7fdda5cdbe691764d1ca1d7d8554922da6b22bef84ee4b7e361ab62dd",
            "sum": -253097.8784024875,
            "top1": 319
          },
          "net_j": {
            "core": 3.262182157533374,
            "dram": 2.3418113710434776,
            "package-0": 31.136468880080244,
            "psys": 49.985781630666295,
            "uncore": 23.804278642838728
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.333976724,
          "watts": {
            "core": 6.217213427181208,
            "dram": 2.2493211058456217,
            "package-0": 29.067808532467314,
            "psys": 47.62766835203041,
            "uncore": 17.846551271609744
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 553.0736012792826,
            "sha256": "7bd1634f8d377902a10467f3d09341ea554283d2d40c1594e26340bce4ce43e1",
            "sum": -121082.53280104871,
            "top1": 42
          },
          "gpu_active_fraction": 0.991869918699187,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 123,
          "gross_j": {
            "core": 9.252662,
            "dram": 4.068776,
            "package-0": 28.523059,
            "psys": 50.331414,
            "uncore": 11.142733
          },
          "net_j": {
            "core": 4.590747533180599,
            "dram": 3.458424214740239,
            "package-0": 21.444798212910484,
            "psys": 37.77802482084503,
            "uncore": 11.14031898685909
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.236006099,
          "token_latencies_s": [
            0.09510276,
            0.088289847,
            0.087852981,
            0.087528247,
            0.08787454,
            0.087843446,
            0.087509492,
            0.087503841,
            0.087877606,
            0.087510847,
            0.087662184,
            0.087578072,
            0.087385574,
            0.088464537
          ],
          "watts": {
            "core": 7.4859355528147775,
            "dram": 3.291873723998509,
            "package-0": 23.076794704392476,
            "psys": 40.72100780143481,
            "uncore": 9.015111664105147
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.996397917158902,
    "compile_target": "NPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 403518,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 2048,
      "NPUW_LLM_MIN_RESPONSE_LEN": 2048,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.043826090171933174,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 26.698051999999997,
            "dram": 9.004066,
            "package-0": 77.026047,
            "psys": 131.9111,
            "uncore": 25.475459
          },
          "net_j": {
            "core": 13.126488455076672,
            "dram": 7.227236309672879,
            "package-0": 56.420120670111004,
            "psys": 95.36621695642731,
            "uncore": 25.471903582146023
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 4.900073,
            "dram": 0.531859,
            "package-0": 6.26512,
            "psys": 10.238316,
            "uncore": 0.0
          },
          "logits": {
            "l2": 887.927456298782,
            "sha256": "a9404b5df4f9d3f7570364ac78827e680855e76bd9d5afb0de3fd418bded0d8e",
            "sum": -221980.5972502306,
            "top1": 8
          },
          "net_j": {
            "core": 3.5355691043787756,
            "dram": 0.35321406890702123,
            "package-0": 4.193371475451029,
            "psys": 6.5640425743245885,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 124142,
          "npu_utilization": 0.3431528777609907,
          "seconds": 0.361768786,
          "watts": {
            "core": 13.54476447285311,
            "dram": 1.4701627685479752,
            "package-0": 17.31802256704369,
            "psys": 28.300716911491637,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 569.172298907028,
          "decode_tokens_per_s": 8.437454835358947,
          "end_to_end_ms": 3598.2074390000002,
          "handoff_estimate_ms": 261.1783375,
          "handoff_plus_first_decode_ms": 361.76878600000003,
          "output_tokens_per_s_e2e": 4.446658585211156,
          "prefill_tokens_per_s": 1116.2258369756726,
          "steady_tpot_ms": 100.5904485,
          "steady_tpot_p95_ms": 103.65122000000001,
          "ttft_ms": 1820.420145
        },
        "prefill": {
          "gpu_active_fraction": 0.9832402234636871,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 179,
          "gross_j": {
            "core": 13.952357,
            "dram": 4.098744,
            "package-0": 48.870175,
            "psys": 82.584322,
            "uncore": 25.475459
          },
          "logits": {
            "l2": 600.6828465575093,
            "sha256": "c54e46c783110f50d983e9def0f6af3c07eb0aa1064735a83c5940836218a748",
            "sum": -153166.4828705974,
            "top1": 90
          },
          "net_j": {
            "core": 7.086174853705449,
            "dram": 3.1998028974257293,
            "package-0": 38.44513863585395,
            "psys": 64.09538204392076,
            "uncore": 25.471903582146023
          },
          "npu_busy_delta_us": 14319,
          "npu_utilization": 0.007865766614003275,
          "seconds": 1.820420145,
          "watts": {
            "core": 7.664360910486408,
            "dram": 2.251537377927665,
            "package-0": 26.84554724041466,
            "psys": 45.365528516495296,
            "uncore": 13.99427438219214
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 861.679016957534,
            "sha256": "ebe03c04ea266bd221813308a6fd828ac7e4973a369c68eb659d36c4bbad5510",
            "sum": -223266.26944584027,
            "top1": 735
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 7.845622,
            "dram": 4.373463,
            "package-0": 21.890752,
            "psys": 39.088462,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.504744496992447,
            "dram": 3.6742193433401287,
            "package-0": 13.781610558806024,
            "psys": 24.706792338181955,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1405989,
          "npu_utilization": 0.9929171067021109,
          "seconds": 1.416018508,
          "token_latencies_s": [
            0.106603858,
            0.101435403,
            0.100607704,
            0.100597683,
            0.100571318,
            0.102061338,
            0.100583214,
            0.100404997,
            0.100654543,
            0.100429012,
            0.100630039,
            0.100432121,
            0.100522984,
            0.100450332
          ],
          "watts": {
            "core": 5.540621083463973,
            "dram": 3.0885634441156613,
            "package-0": 15.459368557914356,
            "psys": 27.604485237420356,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 23.818849,
            "dram": 8.891762,
            "package-0": 73.596186,
            "psys": 124.875839,
            "uncore": 25.280148
          },
          "net_j": {
            "core": 10.902868894360124,
            "dram": 7.2007632543204725,
            "package-0": 53.98564270083212,
            "psys": 90.09628093424675,
            "uncore": 25.276612393139285
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 3.05175,
            "dram": 0.410643,
            "package-0": 4.0902,
            "psys": 6.37126,
            "uncore": 0.0
          },
          "logits": {
            "l2": 778.2465713227663,
            "sha256": "fe566b0d8145c772eb0c4796d094855e007986ea4936bca45001faf4234867f6",
            "sum": -167187.10537097976,
            "top1": 482
          },
          "net_j": {
            "core": 2.299653141681782,
            "dram": 0.31217622987155424,
            "package-0": 2.9482790706872297,
            "psys": 4.3460481197550695,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 120506,
          "npu_utilization": 0.6043361643860966,
          "seconds": 0.199402265,
          "watts": {
            "core": 15.304490147090357,
            "dram": 2.0593697869981566,
            "package-0": 20.51230461198623,
            "psys": 31.951793526517868,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 598.062087386913,
          "decode_tokens_per_s": 9.29300683077449,
          "end_to_end_ms": 3424.3936259999996,
          "handoff_estimate_ms": 98.87701849999999,
          "handoff_plus_first_decode_ms": 199.402265,
          "output_tokens_per_s_e2e": 4.672360057710258,
          "prefill_tokens_per_s": 1122.4803622682073,
          "steady_tpot_ms": 100.5252465,
          "steady_tpot_p95_ms": 103.2223723,
          "ttft_ms": 1810.2766589999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9776536312849162,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 179,
          "gross_j": {
            "core": 13.555874,
            "dram": 4.095449,
            "package-0": 48.221495,
            "psys": 80.659095,
            "uncore": 25.280148
          },
          "logits": {
            "l2": 719.4226608142055,
            "sha256": "0c501c54920f07eb1cd51683351d82fa8c33d783dc082c8bb6ae5bfd66f63919",
            "sum": -181298.67483598553,
            "top1": 264
          },
          "net_j": {
            "core": 6.727950616478253,
            "dram": 3.2015168490756176,
            "package-0": 37.85454753950602,
            "psys": 62.273176483292566,
            "uncore": 25.276612393139285
          },
          "npu_busy_delta_us": 11753,
          "npu_utilization": 0.00649237780400504,
          "seconds": 1.810276659,
          "watts": {
            "core": 7.488288562195951,
            "dram": 2.2623332072691773,
            "package-0": 26.637638374367395,
            "psys": 44.55622548022921,
            "uncore": 13.96479807344188
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 815.9284584349092,
            "sha256": "9eadb231f9593492a54cd405e74b46a6c85686c26e7ef1fd4d66d970c80b3e55",
            "sum": -190088.92014630325,
            "top1": 382
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 7.211225,
            "dram": 4.38567,
            "package-0": 21.284491,
            "psys": 37.845484,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.8752651362000883,
            "dram": 3.687070175373301,
            "package-0": 13.18281609063887,
            "psys": 23.477056331199112,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1403104,
          "npu_utilization": 0.9917929021423288,
          "seconds": 1.414714702,
          "token_latencies_s": [
            0.107573939,
            0.100879221,
            0.100501066,
            0.100397544,
            0.100379416,
            0.100445641,
            0.100631787,
            0.100549427,
            0.100440995,
            0.10065907,
            0.10060175,
            0.100477034,
            0.100652609,
            0.100497828
          ],
          "watts": {
            "core": 5.097299822929245,
            "dram": 3.1000384698059076,
            "package-0": 15.045076558481966,
            "psys": 26.75131879699657,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 24.612486,
            "dram": 8.899635,
            "package-0": 74.324943,
            "psys": 128.319373,
            "uncore": 25.02966
          },
          "net_j": {
            "core": 11.663854247946592,
            "dram": 7.204361403114597,
            "package-0": 54.66482417341999,
            "psys": 93.45189208030197,
            "uncore": 25.026109441274645
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.867302,
            "dram": 0.409362,
            "package-0": 3.889882,
            "psys": 6.263778,
            "uncore": 0.0
          },
          "logits": {
            "l2": 730.2819384422855,
            "sha256": "0079145e0fc0317bb3a1425172a56437e9c22be22e53eb09799308790eff75c3",
            "sum": -132213.13601575233,
            "top1": 66
          },
          "net_j": {
            "core": 2.1146502672167515,
            "dram": 0.3108225840470251,
            "package-0": 2.747118595735032,
            "psys": 4.237071979278087,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 120547,
          "npu_utilization": 0.6040960949524984,
          "seconds": 0.199549378,
          "watts": {
            "core": 14.3688846777563,
            "dram": 2.0514321021837514,
            "package-0": 19.49333061814906,
            "psys": 31.389614253771317,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 596.5539966338031,
          "decode_tokens_per_s": 9.287245410181095,
          "end_to_end_ms": 3433.050506,
          "handoff_estimate_ms": 98.755864,
          "handoff_plus_first_decode_ms": 199.549378,
          "output_tokens_per_s_e2e": 4.660578098701587,
          "prefill_tokens_per_s": 1117.753451453235,
          "steady_tpot_ms": 100.793514,
          "steady_tpot_p95_ms": 102.57646704999999,
          "ttft_ms": 1817.932208
        },
        "prefill": {
          "gpu_active_fraction": 0.9832402234636871,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 179,
          "gross_j": {
            "core": 13.966029,
            "dram": 4.109547,
            "package-0": 48.485227,
            "psys": 81.597874,
            "uncore": 25.02966
          },
          "logits": {
            "l2": 758.2016488784138,
            "sha256": "492b5c2c1dc0d4d194f12400bb823c47d03bde1c1182589d4e78e784bf614b52",
            "sum": -183229.291040726,
            "top1": 264
          },
          "net_j": {
            "core": 7.109230747120069,
            "dram": 3.2118344648177453,
            "package-0": 38.07443835430946,
            "psys": 63.13420256115502,
            "uncore": 25.026109441274645
          },
          "npu_busy_delta_us": 11373,
          "npu_utilization": 0.00625600885992994,
          "seconds": 1.817932208,
          "watts": {
            "core": 7.682370628861206,
            "dram": 2.260561192499649,
            "package-0": 26.67053633058247,
            "psys": 44.88499276316249,
            "uncore": 13.768203176033945
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 746.5391437933649,
            "sha256": "089041255ec886dbd43cd10c181c994eadc76e524ffff1628e8a99374ddb8519",
            "sum": -144414.58760984242,
            "top1": 4194
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 7.779155,
            "dram": 4.380726,
            "package-0": 21.949834,
            "psys": 40.457721,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.439973233609771,
            "dram": 3.6817043542498267,
            "package-0": 13.843267223375499,
            "psys": 26.08061753986886,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1402547,
          "npu_utilization": 0.9908009282939046,
          "seconds": 1.41556892,
          "token_latencies_s": [
            0.104712135,
            0.101426492,
            0.100711233,
            0.100867738,
            0.100934798,
            0.101137329,
            0.100764231,
            0.100919526,
            0.100822797,
            0.100697364,
            0.100594681,
            0.100649415,
            0.100584998,
            0.100717427
          ],
          "watts": {
            "core": 5.495426531404773,
            "dram": 3.094675178372806,
            "package-0": 15.506015772089713,
            "psys": 28.58053778123357,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 7.915970413945615,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 403365,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.047580779530107975,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 18.503737,
            "dram": 7.3125420000000005,
            "package-0": 69.08313600000001,
            "psys": 117.557134,
            "uncore": 35.352082
          },
          "net_j": {
            "core": 8.580078729246372,
            "dram": 6.013306978643074,
            "package-0": 54.01588375796674,
            "psys": 90.83516274969902,
            "uncore": 35.34694337258802
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.504877,
            "dram": 0.309081,
            "package-0": 3.919424,
            "psys": 5.84935,
            "uncore": 0.78076
          },
          "logits": {
            "l2": 529.3319535900215,
            "sha256": "f984afc7983df3c72b7c030f6eb4d656d8481c00cf53a50b1f34deda53645068",
            "sum": -138429.60764685227,
            "top1": 319
          },
          "net_j": {
            "core": 2.0911683236457157,
            "dram": 0.2549170227354055,
            "package-0": 3.291283367656931,
            "psys": 4.73533427576651,
            "uncore": 0.7805457750955458
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.109685935,
          "watts": {
            "core": 22.836811301284893,
            "dram": 2.817872683494014,
            "package-0": 35.73315028950612,
            "psys": 53.328168283381096,
            "uncore": 7.118141446303029
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 778.398228946693,
          "decode_tokens_per_s": 11.11239944154533,
          "end_to_end_ms": 2631.0440129999997,
          "handoff_estimate_ms": 21.822345000000006,
          "handoff_plus_first_decode_ms": 109.685935,
          "output_tokens_per_s_e2e": 6.0812361636460395,
          "prefill_tokens_per_s": 1586.0124603273755,
          "steady_tpot_ms": 87.86358999999999,
          "steady_tpot_p95_ms": 91.53992875,
          "ttft_ms": 1281.200527
        },
        "prefill": {
          "gpu_active_fraction": 0.9133858267716536,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 127,
          "gross_j": {
            "core": 7.011152,
            "dram": 2.925469,
            "package-0": 36.942593,
            "psys": 61.15762,
            "uncore": 23.483094
          },
          "logits": {
            "l2": 813.1669789722381,
            "sha256": "d50d0ae7fdda5cdbe691764d1ca1d7d8554922da6b22bef84ee4b7e361ab62dd",
            "sum": -253097.8784024875,
            "top1": 319
          },
          "net_j": {
            "core": 2.1787751399259925,
            "dram": 2.2927997683814296,
            "package-0": 29.605516380394647,
            "psys": 48.14521751672597,
            "uncore": 23.480591718732278
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.281200527,
          "watts": {
            "core": 5.4723299376226375,
            "dram": 2.283381046408202,
            "package-0": 28.834356700198267,
            "psys": 47.734619765731644,
            "uncore": 18.328976225905034
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 553.0736012792826,
            "sha256": "7bd1634f8d377902a10467f3d09341ea554283d2d40c1594e26340bce4ce43e1",
            "sum": -121082.53280104871,
            "top1": 42
          },
          "gpu_active_fraction": 0.991869918699187,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 123,
          "gross_j": {
            "core": 8.987708,
            "dram": 4.077992,
            "package-0": 28.221119,
            "psys": 50.550164,
            "uncore": 11.088228
          },
          "net_j": {
            "core": 4.310135265674664,
            "dram": 3.4655901875262387,
            "package-0": 21.119084009915163,
            "psys": 37.95461095720654,
            "uncore": 11.085805878760198
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.240157551,
          "token_latencies_s": [
            0.096881593,
            0.088663648,
            0.088054924,
            0.087911311,
            0.087829574,
            0.087763997,
            0.087897606,
            0.087750979,
            0.087544915,
            0.087744268,
            0.087782994,
            0.087769238,
            0.088626289,
            0.08790289
          ],
          "watts": {
            "core": 7.247230799629263,
            "dram": 3.2882854252765825,
            "package-0": 22.75607561091244,
            "psys": 40.76108229896993,
            "uncore": 8.940983338011382
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.156806,
            "dram": 7.409404,
            "package-0": 72.095153,
            "psys": 120.088499,
            "uncore": 35.469391
          },
          "net_j": {
            "core": 11.041868410243916,
            "dram": 6.085126118104263,
            "package-0": 56.73747824749112,
            "psys": 92.85145958393754,
            "uncore": 35.46415332512738
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.407525,
            "dram": 0.306945,
            "package-0": 3.78948,
            "psys": 5.642014,
            "uncore": 0.74951
          },
          "logits": {
            "l2": 529.3319535900215,
            "sha256": "f984afc7983df3c72b7c030f6eb4d656d8481c00cf53a50b1f34deda53645068",
            "sum": -138429.60764685227,
            "top1": 319
          },
          "net_j": {
            "core": 2.0063025474150904,
            "dram": 0.2544157564961882,
            "package-0": 3.18029740489992,
            "psys": 4.5616206057064055,
            "uncore": 0.7493022406599559
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.106375482,
          "watts": {
            "core": 22.632329882180937,
            "dram": 2.8854863379138442,
            "package-0": 35.62362236816939,
            "psys": 53.038669192587065,
            "uncore": 7.04589051826811
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 763.6782683117967,
          "decode_tokens_per_s": 11.122532279262616,
          "end_to_end_ms": 2681.7576,
          "handoff_estimate_ms": 18.484246999999996,
          "handoff_plus_first_decode_ms": 106.37548199999999,
          "output_tokens_per_s_e2e": 5.966236471185912,
          "prefill_tokens_per_s": 1524.2166136629965,
          "steady_tpot_ms": 87.891235,
          "steady_tpot_p95_ms": 91.87286215000002,
          "ttft_ms": 1333.143847
        },
        "prefill": {
          "gpu_active_fraction": 0.8939393939393939,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 132,
          "gross_j": {
            "core": 8.291422,
            "dram": 3.017387,
            "package-0": 38.604699,
            "psys": 62.925742,
            "uncore": 23.638001
          },
          "logits": {
            "l2": 813.1669789722381,
            "sha256": "d50d0ae7fdda5cdbe691764d1ca1d7d8554922da6b22bef84ee4b7e361ab62dd",
            "sum": -253097.8784024875,
            "top1": 319
          },
          "net_j": {
            "core": 3.2631275670608213,
            "dram": 2.3590676537744146,
            "package-0": 30.970157533413136,
            "psys": 49.38578167397346,
            "uncore": 23.63539726951033
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.333143847,
          "watts": {
            "core": 6.219450375635271,
            "dram": 2.2633619071115882,
            "package-0": 28.95763955770633,
            "psys": 47.20101446036978,
            "uncore": 17.731020589558327
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 553.0736012792826,
            "sha256": "7bd1634f8d377902a10467f3d09341ea554283d2d40c1594e26340bce4ce43e1",
            "sum": -121082.53280104871,
            "top1": 42
          },
          "gpu_active_fraction": 0.991869918699187,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 123,
          "gross_j": {
            "core": 10.457859,
            "dram": 4.085072,
            "package-0": 29.700974,
            "psys": 51.520743,
            "uncore": 11.08188
          },
          "net_j": {
            "core": 5.772438295768004,
            "dram": 3.4716427078336607,
            "package-0": 22.58702330917806,
            "psys": 38.90405730425767,
            "uncore": 11.079453814957093
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.242238271,
          "token_latencies_s": [
            0.097536617,
            0.088419048,
            0.088368702,
            0.088131056,
            0.088823148,
            0.08782937,
            0.087889673,
            0.087706404,
            0.087763903,
            0.087746802,
            0.088616436,
            0.087721156,
            0.087755781,
            0.087892797
          ],
          "watts": {
            "core": 8.41856127293634,
            "dram": 3.288477013923845,
            "package-0": 23.909240838386633,
            "psys": 41.47412312335691,
            "uncore": 8.920897269634997
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 20.825264,
            "dram": 7.384259,
            "package-0": 71.662109,
            "psys": 119.028869,
            "uncore": 35.399751
          },
          "net_j": {
            "core": 10.763769832330647,
            "dram": 6.066978090767145,
            "package-0": 56.385578267094395,
            "psys": 91.93573957655872,
            "uncore": 35.39454099897771
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.447198,
            "dram": 0.302307,
            "package-0": 3.768972,
            "psys": 5.542344,
            "uncore": 0.707639
          },
          "logits": {
            "l2": 529.3319535900215,
            "sha256": "f984afc7983df3c72b7c030f6eb4d656d8481c00cf53a50b1f34deda53645068",
            "sum": -138429.60764685227,
            "top1": 319
          },
          "net_j": {
            "core": 2.065541434221394,
            "dram": 0.25233938091624436,
            "package-0": 3.1894966100259246,
            "psys": 4.514636727140632,
            "uncore": 0.707441372186006
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.101188009,
          "watts": {
            "core": 24.184664014883428,
            "dram": 2.987577312643833,
            "package-0": 37.24721967797588,
            "psys": 54.77273497890447,
            "uncore": 6.9933088613296075
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 767.7346817382437,
          "decode_tokens_per_s": 11.219421973091185,
          "end_to_end_ms": 2667.5882289999995,
          "handoff_estimate_ms": 13.265295499999997,
          "handoff_plus_first_decode_ms": 101.188009,
          "output_tokens_per_s_e2e": 5.997927201080029,
          "prefill_tokens_per_s": 1527.1065669282907,
          "steady_tpot_ms": 87.9227135,
          "steady_tpot_p95_ms": 89.9502872,
          "ttft_ms": 1330.620956
        },
        "prefill": {
          "gpu_active_fraction": 0.8939393939393939,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 132,
          "gross_j": {
            "core": 8.981483,
            "dram": 3.007439,
            "package-0": 39.282798,
            "psys": 63.428182,
            "uncore": 23.619691
          },
          "logits": {
            "l2": 813.1669789722381,
            "sha256": "d50d0ae7fdda5cdbe691764d1ca1d7d8554922da6b22bef84ee4b7e361ab62dd",
            "sum": -253097.8784024875,
            "top1": 319
          },
          "net_j": {
            "core": 3.9627042984642706,
            "dram": 2.350365481790241,
            "package-0": 31.66270442363783,
            "psys": 49.91384519848941,
            "uncore": 23.61709219690675
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.330620956,
          "watts": {
            "core": 6.749843341562404,
            "dram": 2.2601770898308327,
            "package-0": 29.522154917872797,
            "psys": 47.66810691954862,
            "uncore": 17.75087856049067
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 553.0736012792826,
            "sha256": "7bd1634f8d377902a10467f3d09341ea554283d2d40c1594e26340bce4ce43e1",
            "sum": -121082.53280104871,
            "top1": 42
          },
          "gpu_active_fraction": 0.991869918699187,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 123,
          "gross_j": {
            "core": 9.396583,
            "dram": 4.074513,
            "package-0": 28.610339,
            "psys": 50.058343,
            "uncore": 11.072421
          },
          "net_j": {
            "core": 4.735524099644983,
            "dram": 3.4642732280606596,
            "package-0": 21.53337723343064,
            "psys": 37.50725765092868,
            "uncore": 11.070007429884955
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.235779264,
          "token_latencies_s": [
            0.092201726,
            0.088244896,
            0.088126783,
            0.087921287,
            0.087696855,
            0.087731693,
            0.087639581,
            0.088042578,
            0.087880844,
            0.087783678,
            0.08792414,
            0.087771034,
            0.088737974,
            0.088048731
          ],
          "watts": {
            "core": 7.603771380323145,
            "dram": 3.297120382819435,
            "package-0": 23.151658094175627,
            "psys": 40.507511703967225,
            "uncore": 8.959869551590081
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.5552231147885323,
    "compile_target": "AUTO:GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 403852,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04400048963725567,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 29.317245999999997,
            "dram": 11.250032,
            "package-0": 125.571579,
            "psys": 207.82448,
            "uncore": 71.998839
          },
          "net_j": {
            "core": 11.522534278419908,
            "dram": 8.920295119422017,
            "package-0": 98.55357789918703,
            "psys": 159.90769749320518,
            "uncore": 71.98962461645857
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.001277,
            "dram": 0.320922,
            "package-0": 3.435172,
            "psys": 5.311571,
            "uncore": 0.771116
          },
          "logits": {
            "l2": 612.9391833447571,
            "sha256": "2baff3d89206006a57e2184111b8b052e1406719a9a1ea9356400754881d45b5",
            "sum": -156671.62222059013,
            "top1": 437
          },
          "net_j": {
            "core": 1.5705418287129977,
            "dram": 0.2645288630944757,
            "package-0": 2.7811797626389487,
            "psys": 4.1517071121486255,
            "uncore": 0.7708929585068237
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.114200143,
          "watts": {
            "core": 17.52429504400883,
            "dram": 2.810171612482131,
            "package-0": 30.08027757023036,
            "psys": 46.51107135653937,
            "uncore": 6.752320791752424
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 868.1858007577679,
          "decode_tokens_per_s": 10.565813551347448,
          "end_to_end_ms": 4717.8841170000005,
          "handoff_estimate_ms": 21.207544999999996,
          "handoff_plus_first_decode_ms": 114.200143,
          "output_tokens_per_s_e2e": 3.3913507842100308,
          "prefill_tokens_per_s": 1237.0342071169125,
          "steady_tpot_ms": 92.99259800000002,
          "steady_tpot_p95_ms": 94.43623765000001,
          "ttft_ms": 3298.211138
        },
        "prefill": {
          "gpu_active_fraction": 0.9296636085626911,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 327,
          "gross_j": {
            "core": 18.231337,
            "dram": 6.70146,
            "package-0": 92.669807,
            "psys": 150.688884,
            "uncore": 59.30124
          },
          "logits": {
            "l2": 664.5119473649122,
            "sha256": "8c551b400c21b7ca89d3bd6971b7ce08a428c386302f3d228b8083472325153d",
            "sum": -177372.0653137702,
            "top1": 72
          },
          "net_j": {
            "core": 5.791286557443822,
            "dram": 5.07277138926368,
            "package-0": 73.78187547236821,
            "psys": 117.19088747406408,
            "uncore": 59.29479834486976
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.298211138,
          "watts": {
            "core": 5.527643997665743,
            "dram": 2.031846876869045,
            "package-0": 28.096990496549587,
            "psys": 45.688064740262845,
            "uncore": 17.979819216776896
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 493.57484335745835,
            "sha256": "2d68b390a1647172ee9cabcff42617290735e761c3331785f10c4fbb6e5d57c9",
            "sum": -72888.27841649903,
            "top1": 42
          },
          "gpu_active_fraction": 0.9923076923076923,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 130,
          "gross_j": {
            "core": 9.084632,
            "dram": 4.22765,
            "package-0": 29.4666,
            "psys": 51.824025,
            "uncore": 11.926483
          },
          "net_j": {
            "core": 4.1607058922630875,
            "dram": 3.582994867063861,
            "package-0": 21.990522664179878,
            "psys": 38.565102906992465,
            "uncore": 11.923933313081985
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.305472836,
          "token_latencies_s": [
            0.095775809,
            0.093471366,
            0.092632614,
            0.093668092,
            0.092907505,
            0.092847205,
            0.09371493,
            0.092613468,
            0.09312234,
            0.092566274,
            0.092846817,
            0.093077691,
            0.092796978,
            0.093400508
          ],
          "watts": {
            "core": 6.9588824443375845,
            "dram": 3.2384051842500376,
            "package-0": 22.571591830502094,
            "psys": 39.697513093255964,
            "uncore": 9.135757306558004
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 33.620459000000004,
            "dram": 11.410677,
            "package-0": 130.722444,
            "psys": 214.424928,
            "uncore": 72.420896
          },
          "net_j": {
            "core": 15.443009817651124,
            "dram": 9.030830986333312,
            "package-0": 103.12332636871619,
            "psys": 165.47752763595605,
            "uncore": 72.41148342894003
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8181818181818182,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.506646,
            "dram": 0.315673,
            "package-0": 3.916799,
            "psys": 5.729844,
            "uncore": 0.760862
          },
          "logits": {
            "l2": 612.9391833447571,
            "sha256": "2baff3d89206006a57e2184111b8b052e1406719a9a1ea9356400754881d45b5",
            "sum": -156671.62222059013,
            "top1": 437
          },
          "net_j": {
            "core": 2.0988071390976213,
            "dram": 0.26227751653032927,
            "package-0": 3.2975706042147643,
            "psys": 4.631634245245796,
            "uncore": 0.7606508145789461
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.10812968,
          "watts": {
            "core": 23.181849793692166,
            "dram": 2.9193927143777727,
            "package-0": 36.22316278009886,
            "psys": 52.990483278966515,
            "uncore": 7.036569422937347
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 849.9056105328265,
          "decode_tokens_per_s": 10.614448028170655,
          "end_to_end_ms": 4819.3587019999995,
          "handoff_estimate_ms": 15.117160500000004,
          "handoff_plus_first_decode_ms": 108.12968000000001,
          "output_tokens_per_s_e2e": 3.3199437911438534,
          "prefill_tokens_per_s": 1197.8190740915454,
          "steady_tpot_ms": 93.0125195,
          "steady_tpot_p95_ms": 94.15325495,
          "ttft_ms": 3406.190541
        },
        "prefill": {
          "gpu_active_fraction": 0.9169139465875371,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 337,
          "gross_j": {
            "core": 20.938179,
            "dram": 6.861128,
            "package-0": 96.219481,
            "psys": 154.843232,
            "uncore": 59.71933
          },
          "logits": {
            "l2": 664.5119473649122,
            "sha256": "8c551b400c21b7ca89d3bd6971b7ce08a428c386302f3d228b8083472325153d",
            "sum": -177372.0653137702,
            "top1": 72
          },
          "net_j": {
            "core": 8.09085650484514,
            "dram": 5.17911811403239,
            "package-0": 76.71318167087709,
            "psys": 120.24855199467956,
            "uncore": 59.71267745295102
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.406190541,
          "watts": {
            "core": 6.147095633074275,
            "dram": 2.0143112716136216,
            "package-0": 28.248414127693394,
            "psys": 45.4593570547996,
            "uncore": 17.532586413227317
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 493.57484335745835,
            "sha256": "2d68b390a1647172ee9cabcff42617290735e761c3331785f10c4fbb6e5d57c9",
            "sum": -72888.27841649903,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 10.175634,
            "dram": 4.233876,
            "package-0": 30.586164,
            "psys": 53.851852,
            "uncore": 11.940704
          },
          "net_j": {
            "core": 5.253346173708362,
            "dram": 3.589435355770593,
            "package-0": 23.112574093624332,
            "psys": 40.59734139603069,
            "uncore": 11.938155161410064
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.305038481,
          "token_latencies_s": [
            0.095537987,
            0.093321664,
            0.093286226,
            0.093186799,
            0.092806638,
            0.093083915,
            0.09297743,
            0.092971573,
            0.092974747,
            0.09340763,
            0.092640208,
            0.092915735,
            0.093047609,
            0.092853653
          ],
          "watts": {
            "core": 7.797190771112596,
            "dram": 3.2442537608207127,
            "package-0": 23.436982468565233,
            "psys": 41.26457019009541,
            "uncore": 9.149694950642608
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 35.532989,
            "dram": 11.419953000000001,
            "package-0": 132.688686,
            "psys": 216.76445999999999,
            "uncore": 72.479001
          },
          "net_j": {
            "core": 17.363264051302657,
            "dram": 9.041118266105288,
            "package-0": 105.10129619867857,
            "psys": 167.83785909759968,
            "uncore": 72.46959242867054
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.61004,
            "dram": 0.316344,
            "package-0": 4.018239,
            "psys": 5.940049,
            "uncore": 0.762755
          },
          "logits": {
            "l2": 612.9391833447571,
            "sha256": "2baff3d89206006a57e2184111b8b052e1406719a9a1ea9356400754881d45b5",
            "sum": -156671.62222059013,
            "top1": 437
          },
          "net_j": {
            "core": 2.1984882480597467,
            "dram": 0.2624624137308097,
            "package-0": 3.39337326118173,
            "psys": 4.8318413427724725,
            "uncore": 0.7625418919851662
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.109114073,
          "watts": {
            "core": 23.920287532479886,
            "dram": 2.89920439501878,
            "package-0": 36.82603801253025,
            "psys": 54.438889839626825,
            "uncore": 6.990436513170945
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 850.2669186723851,
          "decode_tokens_per_s": 10.594658249592095,
          "end_to_end_ms": 4817.310788000001,
          "handoff_estimate_ms": 16.117296000000003,
          "handoff_plus_first_decode_ms": 109.114073,
          "output_tokens_per_s_e2e": 3.3213551510640045,
          "prefill_tokens_per_s": 1199.4697750914913,
          "steady_tpot_ms": 92.99677700000001,
          "steady_tpot_p95_ms": 94.99643845,
          "ttft_ms": 3401.502968
        },
        "prefill": {
          "gpu_active_fraction": 0.9169139465875371,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 337,
          "gross_j": {
            "core": 22.149845,
            "dram": 6.859663,
            "package-0": 97.440119,
            "psys": 156.79214,
            "uncore": 59.743438
          },
          "logits": {
            "l2": 664.5119473649122,
            "sha256": "8c551b400c21b7ca89d3bd6971b7ce08a428c386302f3d228b8083472325153d",
            "sum": -177372.0653137702,
            "top1": 72
          },
          "net_j": {
            "core": 9.320202890403632,
            "dram": 5.17996788298976,
            "package-0": 77.96066408869943,
            "psys": 122.24506892494479,
            "uncore": 59.736794608134666
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.401502968,
          "watts": {
            "core": 6.511781764818968,
            "dram": 2.0166564793660355,
            "package-0": 28.64619549554366,
            "psys": 46.094959044586666,
            "uncore": 17.563835328689326
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 493.57484335745835,
            "sha256": "2d68b390a1647172ee9cabcff42617290735e761c3331785f10c4fbb6e5d57c9",
            "sum": -72888.27841649903,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 10.773104,
            "dram": 4.243946,
            "package-0": 31.230328,
            "psys": 54.032271,
            "uncore": 11.972808
          },
          "net_j": {
            "core": 5.844572912839278,
            "dram": 3.598687969384719,
            "package-0": 23.7472588487974,
            "psys": 40.76094882988241,
            "uncore": 11.970255928550712
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.306693747,
          "token_latencies_s": [
            0.096694802,
            0.093496106,
            0.093174391,
            0.094081935,
            0.092715433,
            0.093077501,
            0.092916285,
            0.092825469,
            0.092931103,
            0.092687669,
            0.093062451,
            0.092663412,
            0.093395732,
            0.092927428
          ],
          "watts": {
            "core": 8.244551582751242,
            "dram": 3.247850546268819,
            "package-0": 23.900265897576077,
            "psys": 41.35037083023556,
            "uncore": 9.16267337123792
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.330430769361556,
    "compile_target": "NPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 403921,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 4096,
      "NPUW_LLM_MIN_RESPONSE_LEN": 4096,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04439142346382141,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 42.042739,
            "dram": 14.471765,
            "package-0": 135.735371,
            "psys": 235.814582,
            "uncore": 55.869975000000004
          },
          "net_j": {
            "core": 19.336425457843905,
            "dram": 11.498986519021546,
            "package-0": 101.26000472961552,
            "psys": 174.67206367467514,
            "uncore": 55.861174704536296
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.046511627906976744,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 43,
          "gross_j": {
            "core": 6.138046,
            "dram": 0.646116,
            "package-0": 7.800151,
            "psys": 12.842374,
            "uncore": 0.009461
          },
          "logits": {
            "l2": 795.940332625534,
            "sha256": "0a1480a30cc1c194375e5006bdf6f9aa21a60cde87b2a188b89fab018ccd6bea",
            "sum": -176995.81349948607,
            "top1": 4194
          },
          "net_j": {
            "core": 4.482238075707148,
            "dram": 0.42933267421562504,
            "package-0": 5.286110816532128,
            "psys": 8.383690463972318,
            "uncore": 0.008603596434036525
          },
          "npu_busy_delta_us": 132967,
          "npu_utilization": 0.30288488496474575,
          "seconds": 0.439001768,
          "watts": {
            "core": 13.981825239482863,
            "dram": 1.4717845054327892,
            "package-0": 17.76792616470738,
            "psys": 29.25358150266037,
            "uncore": 0.021551166053618263
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 680.3885631443926,
          "decode_tokens_per_s": 7.679604594484617,
          "end_to_end_ms": 6020.0894339999995,
          "handoff_estimate_ms": 331.29756,
          "handoff_plus_first_decode_ms": 439.001768,
          "output_tokens_per_s_e2e": 2.6577678247827836,
          "prefill_tokens_per_s": 1003.2300387751951,
          "steady_tpot_ms": 107.704208,
          "steady_tpot_p95_ms": 110.33944955,
          "ttft_ms": 4066.863872
        },
        "prefill": {
          "gpu_active_fraction": 0.9875311720698254,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 401,
          "gross_j": {
            "core": 28.478504,
            "dram": 9.136696,
            "package-0": 105.439611,
            "psys": 180.9313,
            "uncore": 55.860514
          },
          "logits": {
            "l2": 644.3204748970329,
            "sha256": "92de318a3b00199052669a6bdf346331eeba162cc1006f9acf0cb48fca979529",
            "sum": -117752.07326419652,
            "top1": 264
          },
          "net_j": {
            "core": 13.139282345025004,
            "dram": 7.128439223037589,
            "package-0": 82.1498205249381,
            "psys": 139.62654837687626,
            "uncore": 55.85257110810226
          },
          "npu_busy_delta_us": 14168,
          "npu_utilization": 0.003483765487589942,
          "seconds": 4.066863872,
          "watts": {
            "core": 7.002571243181262,
            "dram": 2.2466195790091104,
            "package-0": 25.9265159392087,
            "psys": 44.48914586143295,
            "uncore": 13.735525888779984
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 778.4076845265583,
            "sha256": "6400a338520784bec892978554b291b5e5482e8a25c932a63eee2c6e825f41eb",
            "sum": -170108.64136687852,
            "top1": 4194
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 150,
          "gross_j": {
            "core": 7.426189,
            "dram": 4.688953,
            "package-0": 22.495609,
            "psys": 42.040908,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.7149050371117527,
            "dram": 3.9412146217683333,
            "package-0": 13.824073388145296,
            "psys": 26.66182483382657,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1499090,
          "npu_utilization": 0.9900055764148162,
          "seconds": 1.514223794,
          "token_latencies_s": [
            0.114102581,
            0.108313148,
            0.10751789,
            0.107497365,
            0.107580084,
            0.107286321,
            0.107339312,
            0.107522486,
            0.107883348,
            0.107679407,
            0.10810087,
            0.107893555,
            0.107740762,
            0.107729009
          ],
          "watts": {
            "core": 4.904287615493645,
            "dram": 3.0966050187426917,
            "package-0": 14.856198330218552,
            "psys": 27.763999064460613,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 46.555178,
            "dram": 14.38894,
            "package-0": 139.743783,
            "psys": 240.19658900000002,
            "uncore": 55.433513
          },
          "net_j": {
            "core": 24.289525731451064,
            "dram": 11.473854210930684,
            "package-0": 105.93747993005869,
            "psys": 180.2406631150008,
            "uncore": 55.42547767506553
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 27,
          "gross_j": {
            "core": 4.367969,
            "dram": 0.51721,
            "package-0": 5.643906,
            "psys": 9.042274,
            "uncore": 0.0
          },
          "logits": {
            "l2": 769.6036712572533,
            "sha256": "3cd0382f5a114231fcf2dbc64be60b7477ce00ebf44cb96ae1c6b8dec4e0356c",
            "sum": -160694.08502659574,
            "top1": 4194
          },
          "net_j": {
            "core": 3.3274896555157207,
            "dram": 0.38098733226299286,
            "package-0": 4.064129237415097,
            "psys": 6.240519004890777,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 129842,
          "npu_utilization": 0.4706796488776655,
          "seconds": 0.275860663,
          "watts": {
            "core": 15.833968324798814,
            "dram": 1.8748958056408351,
            "package-0": 20.459263523193954,
            "psys": 32.778410309265446,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 693.8541866602355,
          "decode_tokens_per_s": 8.384259272919909,
          "end_to_end_ms": 5903.257599,
          "handoff_estimate_ms": 168.1576455,
          "handoff_plus_first_decode_ms": 275.860663,
          "output_tokens_per_s_e2e": 2.710367916641545,
          "prefill_tokens_per_s": 991.6895472860117,
          "steady_tpot_ms": 107.7030175,
          "steady_tpot_p95_ms": 110.319164,
          "ttft_ms": 4114.190788
        },
        "prefill": {
          "gpu_active_fraction": 0.9778325123152709,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 406,
          "gross_j": {
            "core": 33.655127,
            "dram": 9.193031,
            "package-0": 110.413841,
            "psys": 187.468453,
            "uncore": 55.433513
          },
          "logits": {
            "l2": 542.5382000105269,
            "sha256": "cfd528695d632340b303a5014801f9770a9431866307b0e28a5e11d8e28e4813",
            "sum": 93871.5732476227,
            "top1": 14924
          },
          "net_j": {
            "core": 18.137399725555284,
            "dram": 7.161403733446306,
            "package-0": 86.85302253084672,
            "psys": 145.68302964258802,
            "uncore": 55.42547767506553
          },
          "npu_busy_delta_us": 14682,
          "npu_utilization": 0.0035686240032483394,
          "seconds": 4.114190788,
          "watts": {
            "core": 8.180254328059615,
            "dram": 2.234468811415753,
            "package-0": 26.837316665539138,
            "psys": 45.56630031519093,
            "uncore": 13.473734169471383
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 752.0471312612783,
            "sha256": "d32a92f79d95feacf0e26e8999682795635ded2922ddc5b9e3abee1402fb5063",
            "sum": -149188.8564177081,
            "top1": 4194
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 150,
          "gross_j": {
            "core": 8.532082,
            "dram": 4.678699,
            "package-0": 23.686036,
            "psys": 43.685862,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.824636350380061,
            "dram": 3.931463145221385,
            "package-0": 15.02032816179687,
            "psys": 28.317114467522025,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1499187,
          "npu_utilization": 0.9907354671942556,
          "seconds": 1.513206148,
          "token_latencies_s": [
            0.112835054,
            0.107687895,
            0.10771814,
            0.108001443,
            0.108964454,
            0.107637014,
            0.107849571,
            0.107833572,
            0.10758187,
            0.107358476,
            0.107729926,
            0.107294444,
            0.107378499,
            0.10729835
          ],
          "watts": {
            "core": 5.638413517733078,
            "dram": 3.0919111756080437,
            "package-0": 15.65288115654682,
            "psys": 28.86973599581225,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 39.387473,
            "dram": 14.329676000000001,
            "package-0": 132.323087,
            "psys": 232.33223,
            "uncore": 55.281902
          },
          "net_j": {
            "core": 17.269484427388246,
            "dram": 11.433922783648349,
            "package-0": 98.74098412851103,
            "psys": 172.7739261340072,
            "uncore": 55.27392373752541
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 3.72637,
            "dram": 0.511229,
            "package-0": 4.987414,
            "psys": 8.372538,
            "uncore": 0.0
          },
          "logits": {
            "l2": 752.0494720512006,
            "sha256": "c4a380935b3e73c9d50bf91e3f1398b1007b698509fefbb1e1e6e91942deba27",
            "sum": -156186.7489903532,
            "top1": 4194
          },
          "net_j": {
            "core": 2.72089169532343,
            "dram": 0.37958877307880623,
            "package-0": 3.460779887144425,
            "psys": 5.6650321974721765,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 132146,
          "npu_utilization": 0.4957069466179144,
          "seconds": 0.266580892,
          "watts": {
            "core": 13.978383717014497,
            "dram": 1.91772559602659,
            "package-0": 18.70882028558896,
            "psys": 31.407119757105473,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 698.486482825295,
          "decode_tokens_per_s": 8.43106963214957,
          "end_to_end_ms": 5864.107754,
          "handoff_estimate_ms": 158.90403450000002,
          "handoff_plus_first_decode_ms": 266.580892,
          "output_tokens_per_s_e2e": 2.7284628235363084,
          "prefill_tokens_per_s": 998.782350409472,
          "steady_tpot_ms": 107.6768575,
          "steady_tpot_p95_ms": 109.48473709999999,
          "ttft_ms": 4084.974067
        },
        "prefill": {
          "gpu_active_fraction": 0.9826302729528535,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 403,
          "gross_j": {
            "core": 28.642078,
            "dram": 9.144142,
            "package-0": 105.18186,
            "psys": 181.811973,
            "uncore": 55.281902
          },
          "logits": {
            "l2": 803.7338034246765,
            "sha256": "5075f102276a647d3b8b6d63b673bd613e0ac84f61524d76ac961b7991a6528e",
            "sum": 218158.74183682352,
            "top1": 7998
          },
          "net_j": {
            "core": 13.234549093033506,
            "dram": 7.126942233307525,
            "package-0": 81.78835750990939,
            "psys": 140.32328674570468,
            "uncore": 55.27392373752541
          },
          "npu_busy_delta_us": 11712,
          "npu_utilization": 0.002867092864704837,
          "seconds": 4.084974067,
          "watts": {
            "core": 7.011569114081233,
            "dram": 2.2384822645191105,
            "package-0": 25.748476801774515,
            "psys": 44.50749748174594,
            "uncore": 13.532987258496592
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 750.6369392275984,
            "sha256": "d113d3583ce066650c8eb822292a21463debdce54ba13c47cac7304db50ab8e8",
            "sum": -154246.8292356953,
            "top1": 4194
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 150,
          "gross_j": {
            "core": 7.019025,
            "dram": 4.674305,
            "package-0": 22.153813,
            "psys": 42.147719,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.3140436390313095,
            "dram": 3.9273917772620184,
            "package-0": 13.49184673145722,
            "psys": 26.785607190830355,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1497198,
          "npu_utilization": 0.9898484237702262,
          "seconds": 1.512552795,
          "token_latencies_s": [
            0.112196164,
            0.107883842,
            0.107862353,
            0.107684435,
            0.107649603,
            0.107772679,
            0.107536952,
            0.107580107,
            0.107524802,
            0.107533906,
            0.107567178,
            0.108010707,
            0.108024738,
            0.10766928
          ],
          "watts": {
            "core": 4.640515705106346,
            "dram": 3.0903417159729623,
            "package-0": 14.646637838515911,
            "psys": 27.865287836118146,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.1715340623632073,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 403768,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04399902559816837,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 30.4326,
            "dram": 11.267245,
            "package-0": 126.691265,
            "psys": 209.48530300000002,
            "uncore": 71.962157
          },
          "net_j": {
            "core": 12.658991993471382,
            "dram": 8.940271080915403,
            "package-0": 99.70530601402503,
            "psys": 161.62534760791368,
            "uncore": 71.95295354429643
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.583794,
            "dram": 0.322448,
            "package-0": 4.039601,
            "psys": 6.001998,
            "uncore": 0.788511
          },
          "logits": {
            "l2": 612.9391833447571,
            "sha256": "2baff3d89206006a57e2184111b8b052e1406719a9a1ea9356400754881d45b5",
            "sum": -156671.62222059013,
            "top1": 437
          },
          "net_j": {
            "core": 2.152811284539373,
            "dram": 0.2660224538708879,
            "package-0": 3.385232912281824,
            "psys": 4.841467536566286,
            "uncore": 0.7882878303245303
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.114265774,
          "watts": {
            "core": 22.61214281014716,
            "dram": 2.821912360213829,
            "package-0": 35.352676996700694,
            "psys": 52.52664721809,
            "uncore": 6.900675262568124
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 869.216651992036,
          "decode_tokens_per_s": 10.593454419911968,
          "end_to_end_ms": 4712.288922,
          "handoff_estimate_ms": 21.5509755,
          "handoff_plus_first_decode_ms": 114.26577400000001,
          "output_tokens_per_s_e2e": 3.3953775468438905,
          "prefill_tokens_per_s": 1237.7438291294693,
          "steady_tpot_ms": 92.7147985,
          "steady_tpot_p95_ms": 94.30396769999999,
          "ttft_ms": 3296.3202109999997
        },
        "prefill": {
          "gpu_active_fraction": 0.9325153374233128,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 326,
          "gross_j": {
            "core": 18.352187,
            "dram": 6.707747,
            "package-0": 92.737617,
            "psys": 150.897746,
            "uncore": 59.233064
          },
          "logits": {
            "l2": 664.5119473649122,
            "sha256": "8c551b400c21b7ca89d3bd6971b7ce08a428c386302f3d228b8083472325153d",
            "sum": -177372.0653137702,
            "top1": 72
          },
          "net_j": {
            "core": 5.919268674303224,
            "dram": 5.079992147334275,
            "package-0": 73.86051428174558,
            "psys": 117.41895451080265,
            "uncore": 59.226626037992816
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.296320211,
          "watts": {
            "core": 5.567477012323546,
            "dram": 2.0349197197577724,
            "package-0": 28.133679698510335,
            "psys": 45.77763576986423,
            "uncore": 17.96945084471346
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 493.57484335745835,
            "sha256": "2d68b390a1647172ee9cabcff42617290735e761c3331785f10c4fbb6e5d57c9",
            "sum": -72888.27841649903,
            "top1": 42
          },
          "gpu_active_fraction": 0.9921875,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 128,
          "gross_j": {
            "core": 9.496619,
            "dram": 4.23705,
            "package-0": 29.914047,
            "psys": 52.585559,
            "uncore": 11.940582
          },
          "net_j": {
            "core": 4.586912034628784,
            "dram": 3.5942564797102396,
            "package-0": 22.459558819997618,
            "psys": 39.36492556054475,
            "uncore": 11.938039675979088
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.301702937,
          "token_latencies_s": [
            0.095749842,
            0.093016564,
            0.092660296,
            0.09352542,
            0.092782285,
            0.092539888,
            0.092575364,
            0.093426931,
            0.092509679,
            0.092391097,
            0.092769301,
            0.092327928,
            0.092529806,
            0.092874187
          ],
          "watts": {
            "core": 7.295534741503008,
            "dram": 3.255005331527496,
            "package-0": 22.980701778965106,
            "psys": 40.397511218029926,
            "uncore": 9.173046830115586
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 33.737644,
            "dram": 11.427095,
            "package-0": 131.091706,
            "psys": 215.956601,
            "uncore": 72.526487
          },
          "net_j": {
            "core": 15.584494342279042,
            "dram": 9.050430352774754,
            "package-0": 103.52948273336894,
            "psys": 167.07463327992537,
            "uncore": 72.51708701161871
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 1.923029,
            "dram": 0.313964,
            "package-0": 3.459403,
            "psys": 5.121142,
            "uncore": 0.75598
          },
          "logits": {
            "l2": 612.9391833447571,
            "sha256": "2baff3d89206006a57e2184111b8b052e1406719a9a1ea9356400754881d45b5",
            "sum": -156671.62222059013,
            "top1": 437
          },
          "net_j": {
            "core": 1.515362010513625,
            "dram": 0.26059101845016064,
            "package-0": 2.8404355593895056,
            "psys": 4.023395052701664,
            "uncore": 0.7557689035766871
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.108084112,
          "watts": {
            "core": 17.791967426257802,
            "dram": 2.9048117636383046,
            "package-0": 32.00658205898014,
            "psys": 47.3810803941286,
            "uncore": 6.994367497787279
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 851.0432810034584,
          "decode_tokens_per_s": 10.610753560155404,
          "end_to_end_ms": 4812.916207,
          "handoff_estimate_ms": 15.01799999999999,
          "handoff_plus_first_decode_ms": 108.08411199999999,
          "output_tokens_per_s_e2e": 3.3243878164197596,
          "prefill_tokens_per_s": 1200.262643236685,
          "steady_tpot_ms": 93.066112,
          "steady_tpot_p95_ms": 94.6056683,
          "ttft_ms": 3399.256007
        },
        "prefill": {
          "gpu_active_fraction": 0.9169139465875371,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 337,
          "gross_j": {
            "core": 21.931828,
            "dram": 6.876203,
            "package-0": 97.316096,
            "psys": 157.912804,
            "uncore": 59.810699
          },
          "logits": {
            "l2": 664.5119473649122,
            "sha256": "8c551b400c21b7ca89d3bd6971b7ce08a428c386302f3d228b8083472325153d",
            "sum": -177372.0653137702,
            "top1": 72
          },
          "net_j": {
            "core": 9.110660880940292,
            "dram": 5.197617454098566,
            "package-0": 77.84950880508796,
            "psys": 123.38855399025039,
            "uncore": 59.80405999661898
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.399256007,
          "watts": {
            "core": 6.451949472130476,
            "dram": 2.022855291228437,
            "package-0": 28.62864573883211,
            "psys": 46.45510772793054,
            "uncore": 17.59523227342494
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 493.57484335745835,
            "sha256": "2d68b390a1647172ee9cabcff42617290735e761c3331785f10c4fbb6e5d57c9",
            "sum": -72888.27841649903,
            "top1": 42
          },
          "gpu_active_fraction": 0.9923076923076923,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 130,
          "gross_j": {
            "core": 9.882787,
            "dram": 4.236928,
            "package-0": 30.316207,
            "psys": 52.922655,
            "uncore": 11.959808
          },
          "net_j": {
            "core": 4.958471450825124,
            "dram": 3.592221880226027,
            "package-0": 22.839538368891464,
            "psys": 39.662684236973305,
            "uncore": 11.957258111423043
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.305576088,
          "token_latencies_s": [
            0.096251376,
            0.093719518,
            0.092846601,
            0.092841851,
            0.092646383,
            0.093118375,
            0.093553867,
            0.093072083,
            0.093112069,
            0.093060141,
            0.093137184,
            0.0929697,
            0.092669052,
            0.092527111
          ],
          "watts": {
            "core": 7.569675249750745,
            "dram": 3.2452555151270506,
            "package-0": 23.220559321395903,
            "psys": 40.53586419545392,
            "uncore": 9.160559932068855
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 32.793555999999995,
            "dram": 11.423739,
            "package-0": 130.21152,
            "psys": 216.322261,
            "uncore": 72.718137
          },
          "net_j": {
            "core": 14.653619383592828,
            "dram": 9.04880424366228,
            "package-0": 102.66935830954654,
            "psys": 167.47587275076995,
            "uncore": 72.70874385354075
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.439202,
            "dram": 0.317321,
            "package-0": 3.854177,
            "psys": 5.788315,
            "uncore": 0.759825
          },
          "logits": {
            "l2": 612.9391833447571,
            "sha256": "2baff3d89206006a57e2184111b8b052e1406719a9a1ea9356400754881d45b5",
            "sum": -156671.62222059013,
            "top1": 437
          },
          "net_j": {
            "core": 2.031070990127524,
            "dram": 0.2638872675133794,
            "package-0": 3.234505029667495,
            "psys": 4.689318559915978,
            "uncore": 0.7596136632995827
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.108207137,
          "watts": {
            "core": 22.541969666936108,
            "dram": 2.9325329991865514,
            "package-0": 35.61851008034711,
            "psys": 53.49291331864737,
            "uncore": 7.021949023565793
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 851.6631767764982,
          "decode_tokens_per_s": 10.629307023993828,
          "end_to_end_ms": 4809.413054000001,
          "handoff_estimate_ms": 15.408518499999996,
          "handoff_plus_first_decode_ms": 108.20713699999999,
          "output_tokens_per_s_e2e": 3.326809284283196,
          "prefill_tokens_per_s": 1200.6284233949712,
          "steady_tpot_ms": 92.7986185,
          "steady_tpot_p95_ms": 94.44744735,
          "ttft_ms": 3398.220399
        },
        "prefill": {
          "gpu_active_fraction": 0.9196428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 336,
          "gross_j": {
            "core": 20.965462,
            "dram": 6.869917,
            "package-0": 96.557798,
            "psys": 157.600976,
            "uncore": 60.018096
          },
          "logits": {
            "l2": 664.5119473649122,
            "sha256": "8c551b400c21b7ca89d3bd6971b7ce08a428c386302f3d228b8083472325153d",
            "sum": -177372.0653137702,
            "top1": 72
          },
          "net_j": {
            "core": 8.148200942505085,
            "dram": 5.1918428473612215,
            "package-0": 77.09714144208331,
            "psys": 123.0872440534473,
            "uncore": 60.011459019239524
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.398220399,
          "watts": {
            "core": 6.169541565393916,
            "dram": 2.021621964844194,
            "package-0": 28.414224700791696,
            "psys": 46.37750277950703,
            "uncore": 17.661625484227457
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 493.57484335745835,
            "sha256": "2d68b390a1647172ee9cabcff42617290735e761c3331785f10c4fbb6e5d57c9",
            "sum": -72888.27841649903,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 9.388892,
            "dram": 4.236501,
            "package-0": 29.799545,
            "psys": 52.93297,
            "uncore": 11.940216
          },
          "net_j": {
            "core": 4.474347450960219,
            "dram": 3.59307412878768,
            "package-0": 22.337711837795737,
            "psys": 39.699310137406684,
            "uncore": 11.937671171001652
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.302985518,
          "token_latencies_s": [
            0.096126099,
            0.093173871,
            0.092844728,
            0.092642769,
            0.093070971,
            0.092752509,
            0.092687359,
            0.093051435,
            0.092973292,
            0.092493251,
            0.093543558,
            0.092651087,
            0.09274077,
            0.092204888
          ],
          "watts": {
            "core": 7.205676402613709,
            "dram": 3.2513799589290597,
            "package-0": 22.87020430260837,
            "psys": 40.62437323267318,
            "uncore": 9.163736538167724
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.562057756818831,
    "compile_target": "AUTO:GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 404334,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04385981895029545,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 59.225496,
            "dram": 21.396856,
            "package-0": 266.189321,
            "psys": 440.650849,
            "uncore": 159.666278
          },
          "net_j": {
            "core": 22.4822087513465,
            "dram": 16.586314938351506,
            "package-0": 210.40138855894233,
            "psys": 341.71021252184516,
            "uncore": 159.64725174396509
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.197016,
            "dram": 0.348571,
            "package-0": 3.797109,
            "psys": 5.899887,
            "uncore": 0.878477
          },
          "logits": {
            "l2": 550.3755398990971,
            "sha256": "27bdcad5725470bbc36ac628654ced5364aca34207ec7bb0e1c27625c40cf33c",
            "sum": -138718.30373605376,
            "top1": 546
          },
          "net_j": {
            "core": 1.7118527744143335,
            "dram": 0.2850519794374833,
            "package-0": 3.060477759885503,
            "psys": 4.593461747126099,
            "uncore": 0.8782257747983395
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.128630568,
          "watts": {
            "core": 17.08004585659608,
            "dram": 2.7098613138363814,
            "package-0": 29.51949182094881,
            "psys": 45.8669124433937,
            "uncore": 6.829457520548304
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 840.9218228464197,
          "decode_tokens_per_s": 9.820576869039325,
          "end_to_end_ms": 9741.690341999998,
          "handoff_estimate_ms": 29.180919499999998,
          "handoff_plus_first_decode_ms": 128.630568,
          "output_tokens_per_s_e2e": 1.6424254352469134,
          "prefill_tokens_per_s": 995.3391975985407,
          "steady_tpot_ms": 99.4496485,
          "steady_tpot_p95_ms": 101.72606005,
          "ttft_ms": 8214.28516
        },
        "prefill": {
          "gpu_active_fraction": 0.9446494464944649,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 813,
          "gross_j": {
            "core": 46.080693,
            "dram": 16.525409,
            "package-0": 229.107324,
            "psys": 377.018688,
            "uncore": 145.504876
          },
          "logits": {
            "l2": 774.371456062169,
            "sha256": "f744ae7d8c64e3ef134b46cd4f17547d69d31eb2f2cfcb5f70b3df06e0cd1022",
            "sum": -244942.2377996092,
            "top1": 27
          },
          "net_j": {
            "core": 15.098406704676524,
            "dram": 12.469115413669758,
            "package-0": 182.06641328844265,
            "psys": 293.59101031664153,
            "uncore": 145.48883288149477
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 8.21428516,
          "watts": {
            "core": 5.609823874193332,
            "dram": 2.0117890574911574,
            "package-0": 27.89132828205833,
            "psys": 45.897930331895125,
            "uncore": 17.713638273546376
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 467.6968416061623,
            "sha256": "26b6c86fb5c84fff5365d961a08fcfa0b1bf1e3e9f2d6b350bbbe31be223463e",
            "sum": -62206.398901291075,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 10.947787,
            "dram": 4.522876,
            "package-0": 33.284888,
            "psys": 57.732274,
            "uncore": 13.282925
          },
          "net_j": {
            "core": 5.671949272255644,
            "dram": 3.8321475452442657,
            "package-0": 25.27449751061418,
            "psys": 43.52574045807751,
            "uncore": 13.280193087671979
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.398774614,
          "token_latencies_s": [
            0.103799705,
            0.099957361,
            0.100360113,
            0.099441587,
            0.099367391,
            0.100609482,
            0.09924028,
            0.099167948,
            0.100404216,
            0.099230005,
            0.099052791,
            0.09945771,
            0.099501575,
            0.099136618
          ],
          "watts": {
            "core": 7.826698376154545,
            "dram": 3.233455879690422,
            "package-0": 23.795747840187786,
            "psys": 41.273464232315554,
            "uncore": 9.496115290522424
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 67.83863099999999,
            "dram": 21.560370000000002,
            "package-0": 275.498013,
            "psys": 444.63826600000004,
            "uncore": 159.881426
          },
          "net_j": {
            "core": 30.44214537873348,
            "dram": 16.664310253765876,
            "package-0": 218.7183188138557,
            "psys": 343.93872694202236,
            "uncore": 159.86206150750317
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.763603,
            "dram": 0.346374,
            "package-0": 4.34087,
            "psys": 6.394942,
            "uncore": 0.866575
          },
          "logits": {
            "l2": 550.3755398990971,
            "sha256": "27bdcad5725470bbc36ac628654ced5364aca34207ec7bb0e1c27625c40cf33c",
            "sum": -138718.30373605376,
            "top1": 546
          },
          "net_j": {
            "core": 2.2975947622579693,
            "dram": 0.28536280773639433,
            "package-0": 3.6333220904132912,
            "psys": 5.140096418704074,
            "uncore": 0.8663336935543994
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.123552036,
          "watts": {
            "core": 22.367927631722715,
            "dram": 2.803466549106483,
            "package-0": 35.13394145928926,
            "psys": 51.759098490291166,
            "uncore": 7.013846376436888
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 826.2335772251334,
          "decode_tokens_per_s": 9.861020473541428,
          "end_to_end_ms": 9914.871806000001,
          "handoff_estimate_ms": 24.143331000000003,
          "handoff_plus_first_decode_ms": 123.552036,
          "output_tokens_per_s_e2e": 1.6137374555178388,
          "prefill_tokens_per_s": 974.0602763729432,
          "steady_tpot_ms": 99.408705,
          "steady_tpot_p95_ms": 101.59664469999998,
          "ttft_ms": 8393.731064000001
        },
        "prefill": {
          "gpu_active_fraction": 0.9314079422382672,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 831,
          "gross_j": {
            "core": 53.764266,
            "dram": 16.680866,
            "package-0": 237.476321,
            "psys": 380.822328,
            "uncore": 145.706231
          },
          "logits": {
            "l2": 774.371456062169,
            "sha256": "f744ae7d8c64e3ef134b46cd4f17547d69d31eb2f2cfcb5f70b3df06e0cd1022",
            "sum": -244942.2377996092,
            "top1": 27
          },
          "net_j": {
            "core": 22.10515338760097,
            "dram": 12.535960287926558,
            "package-0": 189.4077738498242,
            "psys": 295.5721234930767,
            "uncore": 145.68983741010902
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 8.393731064,
          "watts": {
            "core": 6.405288135879212,
            "dram": 1.9873005071061687,
            "package-0": 28.292105047124487,
            "psys": 45.36985103481748,
            "uncore": 17.35893488712328
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 467.6968416061623,
            "sha256": "26b6c86fb5c84fff5365d961a08fcfa0b1bf1e3e9f2d6b350bbbe31be223463e",
            "sum": -62206.398901291075,
            "top1": 42
          },
          "gpu_active_fraction": 0.9927536231884058,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 138,
          "gross_j": {
            "core": 11.310762,
            "dram": 4.53313,
            "package-0": 33.680822,
            "psys": 57.420996,
            "uncore": 13.30862
          },
          "net_j": {
            "core": 6.039397228874539,
            "dram": 3.8429871581029236,
            "package-0": 25.6772228736182,
            "psys": 43.226507030241606,
            "uncore": 13.305890403839753
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.397588706,
          "token_latencies_s": [
            0.10406896,
            0.100265398,
            0.099652442,
            0.099870934,
            0.09952613,
            0.099416567,
            0.099140365,
            0.099119543,
            0.099182989,
            0.099400843,
            0.099159644,
            0.100129151,
            0.099225483,
            0.0994006
          ],
          "watts": {
            "core": 8.093054810361354,
            "dram": 3.2435365143827943,
            "package-0": 24.099237390374274,
            "psys": 41.0857613212567,
            "uncore": 9.52255834843588
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 63.324484,
            "dram": 21.633184,
            "package-0": 271.834753,
            "psys": 448.484143,
            "uncore": 160.576127
          },
          "net_j": {
            "core": 25.868534451965292,
            "dram": 16.72933905862344,
            "package-0": 214.96477376377908,
            "psys": 347.6244822108849,
            "uncore": 160.55673171614018
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.728081,
            "dram": 0.347045,
            "package-0": 4.302235,
            "psys": 6.456588,
            "uncore": 0.861021
          },
          "logits": {
            "l2": 550.3755398990971,
            "sha256": "27bdcad5725470bbc36ac628654ced5364aca34207ec7bb0e1c27625c40cf33c",
            "sum": -138718.30373605376,
            "top1": 546
          },
          "net_j": {
            "core": 2.2568610052231572,
            "dram": 0.2853514689176154,
            "package-0": 3.5867739946853643,
            "psys": 5.187708438709191,
            "uncore": 0.8607769948240691
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.124933821,
          "watts": {
            "core": 21.836208787690886,
            "dram": 2.7778306724485757,
            "package-0": 34.43611157942571,
            "psys": 51.68006508021555,
            "uncore": 6.8918167483246995
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 824.9218739170508,
          "decode_tokens_per_s": 9.825767383491202,
          "end_to_end_ms": 9930.637384000001,
          "handoff_estimate_ms": 25.2402075,
          "handoff_plus_first_decode_ms": 124.933821,
          "output_tokens_per_s_e2e": 1.6111755349942398,
          "prefill_tokens_per_s": 972.865539972189,
          "steady_tpot_ms": 99.6936135,
          "steady_tpot_p95_ms": 102.18888539999999,
          "ttft_ms": 8404.039062
        },
        "prefill": {
          "gpu_active_fraction": 0.9326923076923077,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 832,
          "gross_j": {
            "core": 50.071894,
            "dram": 16.750445,
            "package-0": 234.59144,
            "psys": 384.07232,
            "uncore": 146.354117
          },
          "logits": {
            "l2": 774.371456062169,
            "sha256": "f744ae7d8c64e3ef134b46cd4f17547d69d31eb2f2cfcb5f70b3df06e0cd1022",
            "sum": -244942.2377996092,
            "top1": 27
          },
          "net_j": {
            "core": 18.373902125565756,
            "dram": 12.60044909865739,
            "package-0": 186.4638618320195,
            "psys": 298.71742320178726,
            "uncore": 146.3377032778114
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 8.404039062,
          "watts": {
            "core": 5.958074876925173,
            "dram": 1.993142211313534,
            "package-0": 27.914130130681677,
            "psys": 45.70092037489865,
            "uncore": 17.414735452832428
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 467.6968416061623,
            "sha256": "26b6c86fb5c84fff5365d961a08fcfa0b1bf1e3e9f2d6b350bbbe31be223463e",
            "sum": -62206.398901291075,
            "top1": 42
          },
          "gpu_active_fraction": 0.9927536231884058,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 138,
          "gross_j": {
            "core": 10.524509,
            "dram": 4.535694,
            "package-0": 32.941078,
            "psys": 57.955235,
            "uncore": 13.360989
          },
          "net_j": {
            "core": 5.237771321176379,
            "dram": 3.8435384910484363,
            "package-0": 24.91413793707423,
            "psys": 43.71935057038842,
            "uncore": 13.358251443504688
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.401664501,
          "token_latencies_s": [
            0.104194957,
            0.101108693,
            0.100045986,
            0.099687146,
            0.099022089,
            0.099154508,
            0.099700081,
            0.100846795,
            0.099643308,
            0.099451213,
            0.099772853,
            0.099615578,
            0.099607513,
            0.09978594
          ],
          "watts": {
            "core": 7.508579258796539,
            "dram": 3.23593413171559,
            "package-0": 23.501399925944188,
            "psys": 41.34743724953623,
            "uncore": 9.532230423519872
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 11.38440308533609,
    "compile_target": "NPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 404406,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 8192,
      "NPUW_LLM_MIN_RESPONSE_LEN": 8192,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.044112238101661205,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 86.191186,
            "dram": 28.397510999999998,
            "package-0": 291.347886,
            "psys": 500.813536,
            "uncore": 133.409998
          },
          "net_j": {
            "core": 39.7499536501519,
            "dram": 22.31728596023605,
            "package-0": 220.83540530442116,
            "psys": 375.7587183287195,
            "uncore": 133.3904416045126
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 56,
          "gross_j": {
            "core": 8.534097,
            "dram": 0.848326,
            "package-0": 10.710055,
            "psys": 17.219194,
            "uncore": 0.0
          },
          "logits": {
            "l2": 809.4602388694859,
            "sha256": "02515d7fe1a284c67bb7f8751a7f64870233f7377c943f6424cceca3b74dcbf9",
            "sum": -186452.78511033393,
            "top1": 4194
          },
          "net_j": {
            "core": 6.381426452547956,
            "dram": 0.5664919335311953,
            "package-0": 7.441620190223687,
            "psys": 11.422581610603938,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 149585,
          "npu_utilization": 0.26209224101856354,
          "seconds": 0.570734179,
          "watts": {
            "core": 14.95284024333857,
            "dram": 1.4863767253020954,
            "package-0": 18.765399715092233,
            "psys": 30.170251990462972,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 665.3189531609914,
          "decode_tokens_per_s": 6.522411948421407,
          "end_to_end_ms": 12312.891375,
          "handoff_estimate_ms": 447.624303,
          "handoff_plus_first_decode_ms": 570.734179,
          "output_tokens_per_s_e2e": 1.2994510803925614,
          "prefill_tokens_per_s": 816.5280105048917,
          "steady_tpot_ms": 123.10987600000001,
          "steady_tpot_p95_ms": 125.08208115,
          "ttft_ms": 10013.128631000001
        },
        "prefill": {
          "gpu_active_fraction": 0.9888438133874239,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 986,
          "gross_j": {
            "core": 68.037058,
            "dram": 22.254216,
            "package-0": 253.695578,
            "psys": 436.433208,
            "uncore": 133.409998
          },
          "logits": {
            "l2": 821.0241594431237,
            "sha256": "53efb3fb82e8f96410679387a5bb401865f2ce272df9e8e28dcbcf94fe2e1c8a",
            "sum": 229391.01477363706,
            "top1": 7998
          },
          "net_j": {
            "core": 30.26997152627665,
            "dram": 17.309636081736606,
            "package-0": 196.3531944937754,
            "psys": 334.7357319862653,
            "uncore": 133.3904416045126
          },
          "npu_busy_delta_us": 14349,
          "npu_utilization": 0.0014330186427023838,
          "seconds": 10.013128631,
          "watts": {
            "core": 6.794785177268337,
            "dram": 2.222503756828049,
            "package-0": 25.336294713579814,
            "psys": 43.58609821997401,
            "uncore": 13.323507858170448
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 798.1518897193432,
            "sha256": "7353c5ffa2362e60360074d60553031a1ea2eb55d106d91bd58ec13397d04511",
            "sum": -180498.95023651794,
            "top1": 4194
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 171,
          "gross_j": {
            "core": 9.620031,
            "dram": 5.294969,
            "package-0": 26.942253,
            "psys": 47.161134,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.098555671327297,
            "dram": 4.441157944968246,
            "package-0": 17.040590620422087,
            "psys": 29.60040473185029,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1718527,
          "npu_utilization": 0.9939263207025154,
          "seconds": 1.729028565,
          "token_latencies_s": [
            0.127413598,
            0.123744848,
            0.12322284,
            0.123032849,
            0.123826649,
            0.123082879,
            0.123060194,
            0.122900025,
            0.123136873,
            0.123061888,
            0.123009146,
            0.122924988,
            0.123158256,
            0.123419675
          ],
          "watts": {
            "core": 5.563835783129472,
            "dram": 3.0623953283270366,
            "package-0": 15.582306472767847,
            "psys": 27.276087251918824,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 88.253742,
            "dram": 28.280263,
            "package-0": 292.576887,
            "psys": 505.99620300000004,
            "uncore": 132.704189
          },
          "net_j": {
            "core": 42.37595633406719,
            "dram": 22.273806085300127,
            "package-0": 222.9198965963943,
            "psys": 382.45860868303794,
            "uncore": 132.6845922394805
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 38,
          "gross_j": {
            "core": 5.941391,
            "dram": 0.711058,
            "package-0": 7.685588,
            "psys": 12.708647,
            "uncore": 0.0
          },
          "logits": {
            "l2": 750.7986950958871,
            "sha256": "ba2622eda8d54300ed0e5f25cf093e0b4e612efaabeb35cd823240152f32e4dd",
            "sum": -158572.60981203616,
            "top1": 4194
          },
          "net_j": {
            "core": 4.444766939104243,
            "dram": 0.5151155017126701,
            "package-0": 5.413239273173071,
            "psys": 8.678606460854347,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 150219,
          "npu_utilization": 0.37857839961375367,
          "seconds": 0.396797599,
          "watts": {
            "core": 14.973354211248642,
            "dram": 1.7919916899497166,
            "package-0": 19.369038571223815,
            "psys": 32.02803402043771,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 673.4900484408342,
          "decode_tokens_per_s": 7.043213831919347,
          "end_to_end_ms": 12163.505636,
          "handoff_estimate_ms": 273.27189699999997,
          "handoff_plus_first_decode_ms": 396.797599,
          "output_tokens_per_s_e2e": 1.3154102508610044,
          "prefill_tokens_per_s": 814.8461417421682,
          "steady_tpot_ms": 123.525702,
          "steady_tpot_p95_ms": 124.9987826,
          "ttft_ms": 10033.796052
        },
        "prefill": {
          "gpu_active_fraction": 0.9858156028368794,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 987,
          "gross_j": {
            "core": 74.010553,
            "dram": 22.264409,
            "package-0": 259.182075,
            "psys": 445.078634,
            "uncore": 132.704189
          },
          "logits": {
            "l2": 945.9286264695784,
            "sha256": "8ea9865ceac26fc4449fd72654c4156b870e7e63017d5e7597987edcf6438156",
            "sum": 266038.023974292,
            "top1": 7998
          },
          "net_j": {
            "core": 36.16551403961037,
            "dram": 17.309623309035103,
            "package-0": 201.72133496159248,
            "psys": 343.1712511101162,
            "uncore": 132.6845922394805
          },
          "npu_busy_delta_us": 11663,
          "npu_utilization": 0.0011623716427518235,
          "seconds": 10.033796052,
          "watts": {
            "core": 7.376126903162214,
            "dram": 2.218941752913357,
            "package-0": 25.830909224862925,
            "psys": 44.3579510380106,
            "uncore": 13.2257211839131
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 746.970184871683,
            "sha256": "7f28dff8e147b239d2d5b734afb59f3655057aa1de98d1b34e87ed48de2fcc0b",
            "sum": -156238.34350851923,
            "top1": 4194
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 172,
          "gross_j": {
            "core": 8.301798,
            "dram": 5.304796,
            "package-0": 25.709224,
            "psys": 48.208922,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.765675355352574,
            "dram": 4.4490672745523545,
            "package-0": 15.78532236162875,
            "psys": 30.60875111206738,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1717374,
          "npu_utilization": 0.9910335982816807,
          "seconds": 1.732911985,
          "token_latencies_s": [
            0.127564174,
            0.123385,
            0.123617418,
            0.123434516,
            0.123427281,
            0.1235311,
            0.123520304,
            0.123541957,
            0.123600584,
            0.123533853,
            0.123555496,
            0.123261997,
            0.123441636,
            0.12345531
          ],
          "watts": {
            "core": 4.790663387327198,
            "dram": 3.061203365155328,
            "package-0": 14.835850996783313,
            "psys": 27.81960215942531,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 84.735501,
            "dram": 28.418934,
            "package-0": 289.050957,
            "psys": 504.788808,
            "uncore": 132.352994
          },
          "net_j": {
            "core": 38.52525309884055,
            "dram": 22.368950135106967,
            "package-0": 218.88918376102873,
            "psys": 380.35597464425325,
            "uncore": 132.33324516043228
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 40,
          "gross_j": {
            "core": 5.955795,
            "dram": 0.723326,
            "package-0": 7.741252,
            "psys": 12.8309,
            "uncore": 0.0
          },
          "logits": {
            "l2": 719.7956473364802,
            "sha256": "de7c0de46ead8a45b790320cf84e4976d58ab67f8c2d4bc531c013b20fe8fa29",
            "sum": -141453.3275534436,
            "top1": 67692
          },
          "net_j": {
            "core": 4.411194505066401,
            "dram": 0.5211022834762981,
            "package-0": 5.396059870710108,
            "psys": 8.671670722581343,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 150080,
          "npu_utilization": 0.3664800506359781,
          "seconds": 0.409517516,
          "watts": {
            "core": 14.543443851129435,
            "dram": 1.7662883069450928,
            "package-0": 18.903347714191547,
            "psys": 31.331748945263673,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 668.6445862960204,
          "decode_tokens_per_s": 7.009384025014176,
          "end_to_end_ms": 12251.650828999998,
          "handoff_estimate_ms": 286.1944635,
          "handoff_plus_first_decode_ms": 409.517516,
          "output_tokens_per_s_e2e": 1.3059464576094149,
          "prefill_tokens_per_s": 808.5712907438741,
          "steady_tpot_ms": 123.3230525,
          "steady_tpot_p95_ms": 124.93267545,
          "ttft_ms": 10111.662500999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9819639278557114,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 998,
          "gross_j": {
            "core": 70.39545,
            "dram": 22.395695,
            "package-0": 255.50606,
            "psys": 443.372339,
            "uncore": 132.352994
          },
          "logits": {
            "l2": 979.6350729049458,
            "sha256": "df38e347e9e126abe4db0af89e15a3d83431e738275de280974d22de94ea3543",
            "sum": 273512.2029115502,
            "top1": 7998
          },
          "net_j": {
            "core": 32.25671772744415,
            "dram": 17.402458102202516,
            "package-0": 197.59940061456618,
            "psys": 340.6741122468968,
            "uncore": 132.33324516043228
          },
          "npu_busy_delta_us": 11673,
          "npu_utilization": 0.0011544095739791148,
          "seconds": 10.111662501,
          "watts": {
            "core": 6.9618077139182795,
            "dram": 2.2148380642436556,
            "package-0": 25.268452143723305,
            "psys": 43.84762040427599,
            "uncore": 13.089142758365487
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 716.9232841831466,
            "sha256": "5e3c25a5a0af9c851d52716cf0605f2af73f44412a6d337bfd2ac01c749cd5bb",
            "sum": -139669.67636439204,
            "top1": 67692
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 171,
          "gross_j": {
            "core": 8.384256,
            "dram": 5.299913,
            "package-0": 25.803645,
            "psys": 48.585569,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.8573408663300013,
            "dram": 4.445389749428153,
            "package-0": 15.89372327575246,
            "psys": 31.010191674775072,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1717137,
          "npu_utilization": 0.9922946911860423,
          "seconds": 1.730470812,
          "token_latencies_s": [
            0.127591557,
            0.123425006,
            0.123381172,
            0.123413131,
            0.123231025,
            0.123300249,
            0.12350097,
            0.123387958,
            0.123338739,
            0.123296494,
            0.123307366,
            0.123056069,
            0.12307591,
            0.123127331
          ],
          "watts": {
            "core": 4.8450721860543,
            "dram": 3.0627000254772283,
            "package-0": 14.911343676567022,
            "psys": 28.076503032054607,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.1997664403170347,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 404249,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Llama 3.2",
      "ir": "/data/models/llama32_3b_ov_fp16/openvino_model.xml",
      "key": "llama32_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04428545292466879,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 60.950953999999996,
            "dram": 21.32807,
            "package-0": 267.101367,
            "psys": 432.939077,
            "uncore": 159.02138300000001
          },
          "net_j": {
            "core": 24.204821571830056,
            "dram": 16.5171564389396,
            "package-0": 211.3091146764677,
            "psys": 333.9907791531073,
            "uncore": 159.00235527068608
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.205866,
            "dram": 0.349792,
            "package-0": 3.802603,
            "psys": 5.842942,
            "uncore": 0.873167
          },
          "logits": {
            "l2": 550.3755398990971,
            "sha256": "27bdcad5725470bbc36ac628654ced5364aca34207ec7bb0e1c27625c40cf33c",
            "sum": -138718.30373605376,
            "top1": 546
          },
          "net_j": {
            "core": 1.7176740536662036,
            "dram": 0.2858764502585738,
            "package-0": 3.0613732037645978,
            "psys": 4.528361146988676,
            "uncore": 0.8729142064787708
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.129433568,
          "watts": {
            "core": 17.042456868684944,
            "dram": 2.7024828675046644,
            "package-0": 29.378800714201127,
            "psys": 45.14240077195431,
            "uncore": 6.746062968765568
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 840.856711952097,
          "decode_tokens_per_s": 9.80120117157312,
          "end_to_end_ms": 9742.44468,
          "handoff_estimate_ms": 29.815155999999995,
          "handoff_plus_first_decode_ms": 129.433568,
          "output_tokens_per_s_e2e": 1.6422982655314395,
          "prefill_tokens_per_s": 995.6137446175929,
          "steady_tpot_ms": 99.618412,
          "steady_tpot_p95_ms": 101.63124915,
          "ttft_ms": 8212.020017
        },
        "prefill": {
          "gpu_active_fraction": 0.9446494464944649,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 813,
          "gross_j": {
            "core": 48.994564,
            "dram": 16.444782,
            "package-0": 231.186969,
            "psys": 370.090544,
            "uncore": 144.856624
          },
          "logits": {
            "l2": 774.371456062169,
            "sha256": "f744ae7d8c64e3ef134b46cd4f17547d69d31eb2f2cfcb5f70b3df06e0cd1022",
            "sum": -244942.2377996092,
            "top1": 27
          },
          "net_j": {
            "core": 18.02082127329617,
            "dram": 12.3896069632263,
            "package-0": 184.15903012805333,
            "psys": 286.6858720458596,
            "uncore": 144.84058530548992
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 8.212020017,
          "watts": {
            "core": 5.966201238985605,
            "dram": 2.0025258055821906,
            "package-0": 28.152265644921894,
            "psys": 45.066931550807496,
            "uncore": 17.639584864640742
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 467.6968416061623,
            "sha256": "26b6c86fb5c84fff5365d961a08fcfa0b1bf1e3e9f2d6b350bbbe31be223463e",
            "sum": -62206.398901291075,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 9.750524,
            "dram": 4.533496,
            "package-0": 32.111795,
            "psys": 57.005591,
            "uncore": 13.291592
          },
          "net_j": {
            "core": 4.46632624486768,
            "dram": 3.841673025454729,
            "package-0": 24.088711344649766,
            "psys": 42.776545960259085,
            "uncore": 13.2888557587174
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.400991095,
          "token_latencies_s": [
            0.104067884,
            0.100310782,
            0.100319215,
            0.09950612,
            0.099805906,
            0.100233306,
            0.099607027,
            0.099629797,
            0.100236208,
            0.099498137,
            0.099591541,
            0.09948711,
            0.099322784,
            0.099344216
          ],
          "watts": {
            "core": 6.959733030994034,
            "dram": 3.235920639452744,
            "package-0": 22.92077024229765,
            "psys": 40.689474189698544,
            "uncore": 9.487278004432998
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 62.875570999999994,
            "dram": 21.659246,
            "package-0": 270.850441,
            "psys": 446.837173,
            "uncore": 159.990069
          },
          "net_j": {
            "core": 25.37821603315127,
            "dram": 16.749980137353344,
            "package-0": 213.91759524092666,
            "psys": 345.86601760063456,
            "uncore": 159.97065227575857
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7692307692307693,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.696282,
            "dram": 0.345031,
            "package-0": 4.268483,
            "psys": 6.349837,
            "uncore": 0.849973
          },
          "logits": {
            "l2": 550.3755398990971,
            "sha256": "27bdcad5725470bbc36ac628654ced5364aca34207ec7bb0e1c27625c40cf33c",
            "sum": -138718.30373605376,
            "top1": 546
          },
          "net_j": {
            "core": 2.2263685078491418,
            "dram": 0.2835085201490122,
            "package-0": 3.5550056789360203,
            "psys": 5.08447552897968,
            "uncore": 0.8497296713518196
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.12458743,
          "watts": {
            "core": 21.641685682094895,
            "dram": 2.769388533016533,
            "package-0": 34.260944302326486,
            "psys": 50.966915362167754,
            "uncore": 6.822301415158816
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 824.0109767162062,
          "decode_tokens_per_s": 9.885514147172476,
          "end_to_end_ms": 9941.615138,
          "handoff_estimate_ms": 25.80042099999999,
          "handoff_plus_first_decode_ms": 124.58743,
          "output_tokens_per_s_e2e": 1.6093964388988402,
          "prefill_tokens_per_s": 970.5322642211062,
          "steady_tpot_ms": 98.78700900000001,
          "steady_tpot_p95_ms": 102.62800239999999,
          "ttft_ms": 8424.243378
        },
        "prefill": {
          "gpu_active_fraction": 0.9303721488595438,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 833,
          "gross_j": {
            "core": 49.195308,
            "dram": 16.786639,
            "package-0": 233.241773,
            "psys": 382.295163,
            "uncore": 145.848076
          },
          "logits": {
            "l2": 774.371456062169,
            "sha256": "f744ae7d8c64e3ef134b46cd4f17547d69d31eb2f2cfcb5f70b3df06e0cd1022",
            "sum": -244942.2377996092,
            "top1": 27
          },
          "net_j": {
            "core": 17.421110358354095,
            "dram": 12.626666011691132,
            "package-0": 184.99849037247964,
            "psys": 296.7350628115679,
            "uncore": 145.83162281725828
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 8.424243378,
          "watts": {
            "core": 5.839730144605515,
            "dram": 1.9926583607304704,
            "package-0": 27.68696991935363,
            "psys": 45.380355937765025,
            "uncore": 17.31289914781947
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 467.6968416061623,
            "sha256": "26b6c86fb5c84fff5365d961a08fcfa0b1bf1e3e9f2d6b350bbbe31be223463e",
            "sum": -62206.398901291075,
            "top1": 42
          },
          "gpu_active_fraction": 0.9927536231884058,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 138,
          "gross_j": {
            "core": 10.983981,
            "dram": 4.527576,
            "package-0": 33.340185,
            "psys": 58.192173,
            "uncore": 13.29202
          },
          "net_j": {
            "core": 5.730737166948033,
            "dram": 3.8398056055132006,
            "package-0": 25.364099189511013,
            "psys": 44.04647926008698,
            "uncore": 13.289299787148465
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.39278433,
          "token_latencies_s": [
            0.108252234,
            0.09959957,
            0.098821045,
            0.098817327,
            0.098861228,
            0.098751283,
            0.098757316,
            0.098673786,
            0.098694255,
            0.098816977,
            0.098816702,
            0.098500365,
            0.098690247,
            0.09869122
          ],
          "watts": {
            "core": 7.886347342807913,
            "dram": 3.250737319826107,
            "package-0": 23.937794446610408,
            "psys": 41.7811801486882,
            "uncore": 9.543487612328322
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 60.341703,
            "dram": 21.691596,
            "package-0": 267.946579,
            "psys": 441.157135,
            "uncore": 159.68068200000002
          },
          "net_j": {
            "core": 22.860696204017362,
            "dram": 16.78447048877868,
            "package-0": 211.03855493544845,
            "psys": 340.2300012043576,
            "uncore": 159.66127374110036
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.650994,
            "dram": 0.343933,
            "package-0": 4.21465,
            "psys": 6.277328,
            "uncore": 0.859007
          },
          "logits": {
            "l2": 550.3755398990971,
            "sha256": "27bdcad5725470bbc36ac628654ced5364aca34207ec7bb0e1c27625c40cf33c",
            "sum": -138718.30373605376,
            "top1": 546
          },
          "net_j": {
            "core": 2.186928635699177,
            "dram": 0.28317617453823396,
            "package-0": 3.510051986854133,
            "psys": 5.027714099080647,
            "uncore": 0.8587666996050363
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.123036925,
          "watts": {
            "core": 21.546328470091396,
            "dram": 2.795364074646696,
            "package-0": 34.255163642947025,
            "psys": 51.01987066077927,
            "uncore": 6.981700818676994
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 824.3703873456008,
          "decode_tokens_per_s": 9.83424115114686,
          "end_to_end_ms": 9937.280772999999,
          "handoff_estimate_ms": 23.322992,
          "handoff_plus_first_decode_ms": 123.03692500000001,
          "output_tokens_per_s_e2e": 1.6100984127843765,
          "prefill_tokens_per_s": 971.9450880806507,
          "steady_tpot_ms": 99.71393300000001,
          "steady_tpot_p95_ms": 102.0876229,
          "ttft_ms": 8411.997859
        },
        "prefill": {
          "gpu_active_fraction": 0.9326923076923077,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 832,
          "gross_j": {
            "core": 46.987917,
            "dram": 16.807818,
            "package-0": 230.63802,
            "psys": 377.466868,
            "uncore": 145.514764
          },
          "logits": {
            "l2": 774.371456062169,
            "sha256": "f744ae7d8c64e3ef134b46cd4f17547d69d31eb2f2cfcb5f70b3df06e0cd1022",
            "sum": -244942.2377996092,
            "top1": 27
          },
          "net_j": {
            "core": 15.259906478487245,
            "dram": 12.653891967599463,
            "package-0": 182.46486403045114,
            "psys": 292.03113836753204,
            "uncore": 145.49833473368057
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 8.411997859,
          "watts": {
            "core": 5.585821321831129,
            "dram": 1.998076828088741,
            "package-0": 27.4177459226574,
            "psys": 44.87243985638299,
            "uncore": 17.298478487404
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 467.6968416061623,
            "sha256": "26b6c86fb5c84fff5365d961a08fcfa0b1bf1e3e9f2d6b350bbbe31be223463e",
            "sum": -62206.398901291075,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 10.702792,
            "dram": 4.539845,
            "package-0": 33.093909,
            "psys": 57.412939,
            "uncore": 13.306911
          },
          "net_j": {
            "core": 5.413861089830939,
            "dram": 3.8474023466409815,
            "package-0": 25.063638918143177,
            "psys": 43.17114873774494,
            "uncore": 13.304172307814763
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.402245989,
          "token_latencies_s": [
            0.103922517,
            0.101099603,
            0.100158725,
            0.100695298,
            0.100046673,
            0.099575185,
            0.099419181,
            0.09966279,
            0.099392544,
            0.099364709,
            0.099648997,
            0.099564595,
            0.099765076,
            0.09990528
          ],
          "watts": {
            "core": 7.6326066067998575,
            "dram": 3.237552494792695,
            "package-0": 23.600644437286384,
            "psys": 40.94355730048731,
            "uncore": 9.48971229326868
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 0.9332987340167165,
    "compile_target": "AUTO:GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 392871,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.038815573789179325,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 6.166489,
            "dram": 0.9864470000000001,
            "package-0": 11.843659,
            "psys": 18.605971,
            "uncore": 3.52825
          },
          "net_j": {
            "core": 5.104705244065135,
            "dram": 0.7743155594658269,
            "package-0": 9.986823504106466,
            "psys": 14.313075671403713,
            "uncore": 3.5273319524506084
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 4,
          "gross_j": {
            "core": 1.332578,
            "dram": 0.065063,
            "package-0": 1.628414,
            "psys": 2.174006,
            "uncore": 0.145324
          },
          "logits": {
            "l2": 712.6588257239783,
            "sha256": "990e8b5de951b6913e147a016df0e6896d435e87a38f418db97e9b4ce13bdab3",
            "sum": 24349.479562914465,
            "top1": 78
          },
          "net_j": {
            "core": 1.2409058243885105,
            "dram": 0.04674801772543606,
            "package-0": 1.46809872986166,
            "psys": 1.8033664491073382,
            "uncore": 0.14524473770605634
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.037317903,
          "watts": {
            "core": 35.708812469982576,
            "dram": 1.7434795304548596,
            "package-0": 43.63626755769208,
            "psys": 58.25638166217432,
            "uncore": 3.894216671285094
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2369.104330887212,
          "decode_tokens_per_s": 59.777697502971996,
          "end_to_end_ms": 432.230859,
          "handoff_estimate_ms": 22.103882,
          "handoff_plus_first_decode_ms": 37.317903,
          "output_tokens_per_s_e2e": 37.01725517011269,
          "prefill_tokens_per_s": 5559.81017087183,
          "steady_tpot_ms": 15.214021,
          "steady_tpot_p95_ms": 15.5608924,
          "ttft_ms": 181.301154
        },
        "prefill": {
          "gpu_active_fraction": 0.6111111111111112,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 18,
          "gross_j": {
            "core": 1.311703,
            "dram": 0.286376,
            "package-0": 3.858328,
            "psys": 6.459456,
            "uncore": 1.851985
          },
          "logits": {
            "l2": 838.9345161742121,
            "sha256": "e1f6e0509b78af0035fc75a4b5671053a3706868fabf97ca68685fc192833e98",
            "sum": 177352.29018324683,
            "top1": 291
          },
          "net_j": {
            "core": 0.8663330865819363,
            "dram": 0.1973965248706503,
            "package-0": 3.0794701022263014,
            "psys": 4.658781983164668,
            "uncore": 1.851599920876967
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.181301154,
          "watts": {
            "core": 7.234940159288783,
            "dram": 1.5795597197357059,
            "package-0": 21.281320691428142,
            "psys": 35.62832258640781,
            "uncore": 10.214965316767923
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 720.1119452755735,
            "sha256": "9d31e7729784e25aa0cb77989028372b1f68cbebfb778541584dec15714a8c90",
            "sum": 72014.2167466206,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.522208,
            "dram": 0.635008,
            "package-0": 6.356917,
            "psys": 9.972509,
            "uncore": 1.530941
          },
          "net_j": {
            "core": 2.9974663330946885,
            "dram": 0.5301710168697406,
            "package-0": 5.439254672018505,
            "psys": 7.850927239131707,
            "uncore": 1.5304872938675853
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.213611802,
          "token_latencies_s": [
            0.015784664,
            0.015339931,
            0.015102557,
            0.015052076,
            0.01507876,
            0.015395917,
            0.01517187,
            0.015128386,
            0.015362293,
            0.0154404,
            0.015105483,
            0.015256172,
            0.015288043,
            0.015078236
          ],
          "watts": {
            "core": 16.48882677371918,
            "dram": 2.972719644020418,
            "package-0": 29.759203098712685,
            "psys": 46.685196728971,
            "uncore": 7.166930785968465
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.070908,
            "dram": 1.028623,
            "package-0": 11.842437,
            "psys": 18.839551,
            "uncore": 3.545096
          },
          "net_j": {
            "core": 4.964701560495243,
            "dram": 0.8076164492511504,
            "package-0": 9.907915614475579,
            "psys": 14.367050427195585,
            "uncore": 3.5441395433719607
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 2100,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.819212,
            "dram": 0.053833,
            "package-0": 1.002744,
            "psys": 1.362789,
            "uncore": 0.065002
          },
          "logits": {
            "l2": 712.6588257239783,
            "sha256": "990e8b5de951b6913e147a016df0e6896d435e87a38f418db97e9b4ce13bdab3",
            "sum": 24349.479562914465,
            "top1": 78
          },
          "net_j": {
            "core": 0.7597050110648914,
            "dram": 0.04194422960386476,
            "package-0": 0.8986788467229406,
            "psys": 1.122196437352027,
            "uncore": 0.06495054861448182
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.024224101,
          "watts": {
            "core": 33.818055827954154,
            "dram": 2.222290932489094,
            "package-0": 41.39447734303948,
            "psys": 56.257567618298815,
            "uncore": 2.683360674561256
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2273.966598655079,
          "decode_tokens_per_s": 61.96977063360149,
          "end_to_end_ms": 450.31444200000004,
          "handoff_estimate_ms": 8.991238500000001,
          "handoff_plus_first_decode_ms": 24.224101,
          "output_tokens_per_s_e2e": 35.53072810398561,
          "prefill_tokens_per_s": 4840.081870368971,
          "steady_tpot_ms": 15.2328625,
          "steady_tpot_p95_ms": 16.608904,
          "ttft_ms": 208.26094
        },
        "prefill": {
          "gpu_active_fraction": 0.5714285714285714,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 1.810421,
            "dram": 0.334838,
            "package-0": 4.549976,
            "psys": 7.603801,
            "uncore": 1.94012
          },
          "logits": {
            "l2": 838.9345161742121,
            "sha256": "e1f6e0509b78af0035fc75a4b5671053a3706868fabf97ca68685fc192833e98",
            "sum": 177352.29018324683,
            "top1": 291
          },
          "net_j": {
            "core": 1.2988238326914865,
            "dram": 0.23262712752455514,
            "package-0": 3.6553006251440054,
            "psys": 5.535363733573653,
            "uncore": 1.9396776589648335
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.20826094,
          "watts": {
            "core": 8.693041527614348,
            "dram": 1.6077810846335372,
            "package-0": 21.847476535926514,
            "psys": 36.51093191070779,
            "uncore": 9.315813133274055
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 720.1119452755735,
            "sha256": "9d31e7729784e25aa0cb77989028372b1f68cbebfb778541584dec15714a8c90",
            "sum": 72014.2167466206,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2100,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.441275,
            "dram": 0.639952,
            "package-0": 6.289717,
            "psys": 9.872961,
            "uncore": 1.539974
          },
          "net_j": {
            "core": 2.906172716738865,
            "dram": 0.5330450921227304,
            "package-0": 5.353936142608633,
            "psys": 7.7094902562699055,
            "uncore": 1.5395113357926453
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.217829401,
          "token_latencies_s": [
            0.017152317,
            0.016282604,
            0.016316297,
            0.016097107,
            0.016078418,
            0.015614612,
            0.015297622,
            0.015066855,
            0.014909969,
            0.014972133,
            0.015168103,
            0.014985535,
            0.014934098,
            0.014917102
          ],
          "watts": {
            "core": 15.798028109162363,
            "dram": 2.9378586961270665,
            "package-0": 28.874509001656758,
            "psys": 45.324281087289954,
            "uncore": 7.06963335954819
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.166732,
            "dram": 1.03192,
            "package-0": 11.956938999999998,
            "psys": 18.926952999999997,
            "uncore": 3.551992
          },
          "net_j": {
            "core": 5.054146837800091,
            "dram": 0.8096390582779157,
            "package-0": 10.01126257593537,
            "psys": 14.428662628936411,
            "uncore": 3.551030028152213
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2100,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.806272,
            "dram": 0.052979,
            "package-0": 0.982724,
            "psys": 1.34924,
            "uncore": 0.065185
          },
          "logits": {
            "l2": 712.6588257239783,
            "sha256": "990e8b5de951b6913e147a016df0e6896d435e87a38f418db97e9b4ce13bdab3",
            "sum": 24349.479562914465,
            "top1": 78
          },
          "net_j": {
            "core": 0.749675326477962,
            "dram": 0.041671675117165075,
            "package-0": 0.8837483735997929,
            "psys": 1.1204141265405545,
            "uncore": 0.06513606495353673
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.02303937,
          "watts": {
            "core": 34.99540134995011,
            "dram": 2.299498640804848,
            "package-0": 42.65411771242009,
            "psys": 58.56236520356242,
            "uncore": 2.8292874327726847
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2260.9293922967113,
          "decode_tokens_per_s": 61.171514515309156,
          "end_to_end_ms": 452.911092,
          "handoff_estimate_ms": 7.680473,
          "handoff_plus_first_decode_ms": 23.03937,
          "output_tokens_per_s_e2e": 35.327021754636114,
          "prefill_tokens_per_s": 4853.1788458658275,
          "steady_tpot_ms": 15.358897,
          "steady_tpot_p95_ms": 17.65771065,
          "ttft_ms": 207.69891900000002
        },
        "prefill": {
          "gpu_active_fraction": 0.5714285714285714,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 1.851802,
            "dram": 0.336242,
            "package-0": 4.592457,
            "psys": 7.615214,
            "uncore": 1.932734
          },
          "logits": {
            "l2": 838.9345161742121,
            "sha256": "e1f6e0509b78af0035fc75a4b5671053a3706868fabf97ca68685fc192833e98",
            "sum": 177352.29018324683,
            "top1": 291
          },
          "net_j": {
            "core": 1.3415854485264476,
            "dram": 0.23430695772468538,
            "package-0": 3.7001960305954156,
            "psys": 5.552358698107733,
            "uncore": 1.9322928526834007
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.207698919,
          "watts": {
            "core": 8.91579989398019,
            "dram": 1.6188914300512078,
            "package-0": 22.11112615371869,
            "psys": 36.66467806700525,
            "uncore": 9.305460082823059
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 720.1119452755735,
            "sha256": "9d31e7729784e25aa0cb77989028372b1f68cbebfb778541584dec15714a8c90",
            "sum": 72014.2167466206,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.508658,
            "dram": 0.642699,
            "package-0": 6.381758,
            "psys": 9.962499,
            "uncore": 1.554073
          },
          "net_j": {
            "core": 2.9628860627956812,
            "dram": 0.5336604254360653,
            "package-0": 5.427318171740163,
            "psys": 7.755889804288123,
            "uncore": 1.5536011105152754
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.222172803,
          "token_latencies_s": [
            0.018899561,
            0.016467577,
            0.016190065,
            0.016147038,
            0.016989022,
            0.015474793,
            0.015108714,
            0.01518469,
            0.015155258,
            0.015127353,
            0.015147554,
            0.015243001,
            0.015802612,
            0.015202327
          ],
          "watts": {
            "core": 15.792473032804109,
            "dram": 2.892788817180292,
            "package-0": 28.7242988962965,
            "psys": 44.84121758143367,
            "uncore": 6.9948840677857405
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.1501153903082013,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 392939,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.039530511014163494,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 7.110149,
            "dram": 1.022093,
            "package-0": 11.441499,
            "psys": 18.053055,
            "uncore": 1.5514480000000002
          },
          "net_j": {
            "core": 5.822267380230705,
            "dram": 0.7647899646954137,
            "package-0": 9.189266101606341,
            "psys": 12.846023946464287,
            "uncore": 1.5508980303238797
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.07142857142857142,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 14,
          "gross_j": {
            "core": 3.039298,
            "dram": 0.125183,
            "package-0": 3.423453,
            "psys": 4.859484,
            "uncore": 0.006652
          },
          "logits": {
            "l2": 714.5644081016375,
            "sha256": "f7f64affdcb433a14d202661deba5ad255d97611c0d28dc97208fa573f8d89b8",
            "sum": 24380.198635727167,
            "top1": 78
          },
          "net_j": {
            "core": 2.6901048271610217,
            "dram": 0.05541846426945424,
            "package-0": 2.8127879049400257,
            "psys": 3.447661929753406,
            "uncore": 0.0063500779421447745
          },
          "npu_busy_delta_us": 27100,
          "npu_utilization": 0.19064431256344683,
          "seconds": 0.142149533,
          "watts": {
            "core": 21.380991803891472,
            "dram": 0.8806430619789655,
            "package-0": 24.083462870046855,
            "psys": 34.18571906247487,
            "uncore": 0.04679579214657005
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1953.1892186656023,
          "decode_tokens_per_s": 36.8110678885641,
          "end_to_end_ms": 524.270762,
          "handoff_estimate_ms": 124.520972,
          "handoff_plus_first_decode_ms": 142.149533,
          "output_tokens_per_s_e2e": 30.518581541650036,
          "prefill_tokens_per_s": 8631.273461939822,
          "steady_tpot_ms": 17.628561,
          "steady_tpot_p95_ms": 23.970691000000002,
          "ttft_ms": 116.784621
        },
        "prefill": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 0.723204,
            "dram": 0.216858,
            "package-0": 2.811882,
            "psys": 4.684071,
            "uncore": 1.544796
          },
          "logits": {
            "l2": 840.9680755108035,
            "sha256": "8fb49fad0852bc3258a6ca67a848f5fb785395020a78359d88430d551db8ad9a",
            "sum": 178291.6503057778,
            "top1": 291
          },
          "net_j": {
            "core": 0.4363202411501727,
            "dram": 0.1595421250014325,
            "package-0": 2.310182907646963,
            "psys": 3.524171970340973,
            "uncore": 1.544547952381735
          },
          "npu_busy_delta_us": 5719,
          "npu_utilization": 0.04897048901670024,
          "seconds": 0.116784621,
          "watts": {
            "core": 6.192630449175324,
            "dram": 1.8569054567553034,
            "package-0": 24.07750246498638,
            "psys": 40.108628686648736,
            "uncore": 13.227734840189274
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 723.4091300558687,
            "sha256": "1c462e5b54f0c0f9e4e380bd8290810c4daf6d05e264abf4482ad7bacb911518",
            "sum": 73007.79722458497,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 27,
          "gross_j": {
            "core": 3.347647,
            "dram": 0.680052,
            "package-0": 5.206164,
            "psys": 8.5095,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.6958423119195105,
            "dram": 0.5498293754245269,
            "package-0": 4.066295289019352,
            "psys": 5.874190046369908,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 248723,
          "npu_utilization": 0.9373866722529295,
          "seconds": 0.265336608,
          "token_latencies_s": [
            0.026586083,
            0.022562403,
            0.022048623,
            0.019552663,
            0.018827059,
            0.01810178,
            0.017711709,
            0.017545413,
            0.017389837,
            0.017190402,
            0.017081727,
            0.017024312,
            0.016983586,
            0.016690714
          ],
          "watts": {
            "core": 12.616604339797696,
            "dram": 2.5629784187186115,
            "package-0": 19.62097894912413,
            "psys": 32.07058409369581,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 4.605396,
            "dram": 0.955686,
            "package-0": 8.691201999999999,
            "psys": 14.287378,
            "uncore": 1.554684
          },
          "net_j": {
            "core": 3.609645936461058,
            "dram": 0.7567472855867913,
            "package-0": 6.949845483023134,
            "psys": 10.261463332985626,
            "uncore": 1.5543844131486093
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.3333333333333333,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.82379,
            "dram": 0.055908,
            "package-0": 0.979185,
            "psys": 1.360592,
            "uncore": 0.016113
          },
          "logits": {
            "l2": 714.5644081016375,
            "sha256": "f7f64affdcb433a14d202661deba5ad255d97611c0d28dc97208fa573f8d89b8",
            "sum": 24380.198635727167,
            "top1": 78
          },
          "net_j": {
            "core": 0.7620824579795547,
            "dram": 0.04357958594445183,
            "package-0": 0.8712715442257398,
            "psys": 1.1111023865094543,
            "uncore": 0.016059645955530773
          },
          "npu_busy_delta_us": 21826,
          "npu_utilization": 0.868872816462421,
          "seconds": 0.025119902,
          "watts": {
            "core": 32.79431583769714,
            "dram": 2.2256456255283164,
            "package-0": 38.980446659385855,
            "psys": 54.16390557574628,
            "uncore": 0.6414435852496558
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2526.212738275766,
          "decode_tokens_per_s": 51.827849389430625,
          "end_to_end_ms": 405.34986799999996,
          "handoff_estimate_ms": 7.256595499999998,
          "handoff_plus_first_decode_ms": 25.119902,
          "output_tokens_per_s_e2e": 39.47207403555884,
          "prefill_tokens_per_s": 8694.888595877277,
          "steady_tpot_ms": 17.863306500000004,
          "steady_tpot_p95_ms": 23.0763647,
          "ttft_ms": 115.93018
        },
        "prefill": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 0.625914,
            "dram": 0.218139,
            "package-0": 2.71148,
            "psys": 4.606617,
            "uncore": 1.538571
          },
          "logits": {
            "l2": 840.9680755108035,
            "sha256": "8fb49fad0852bc3258a6ca67a848f5fb785395020a78359d88430d551db8ad9a",
            "sum": 178291.6503057778,
            "top1": 291
          },
          "net_j": {
            "core": 0.34112919256771773,
            "dram": 0.1612424698871744,
            "package-0": 2.2134515302897273,
            "psys": 3.4552042357268995,
            "uncore": 1.5383247671930786
          },
          "npu_busy_delta_us": 5583,
          "npu_utilization": 0.048158296657522665,
          "seconds": 0.11593018,
          "watts": {
            "core": 5.399060020436439,
            "dram": 1.8816411740238823,
            "package-0": 23.388905287648136,
            "psys": 39.73613255840714,
            "uncore": 13.271531192309025
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 723.4091300558687,
            "sha256": "1c462e5b54f0c0f9e4e380bd8290810c4daf6d05e264abf4482ad7bacb911518",
            "sum": 73007.79722458497,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 3.155692,
            "dram": 0.681639,
            "package-0": 5.000537,
            "psys": 8.320169,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.5064342859137856,
            "dram": 0.551925229755165,
            "package-0": 3.865122408507667,
            "psys": 5.695156710749272,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 251347,
          "npu_utilization": 0.9509920677726164,
          "seconds": 0.264299786,
          "token_latencies_s": [
            0.023258002,
            0.02297856,
            0.022379059,
            0.020607237,
            0.019030997,
            0.018320396,
            0.018006654,
            0.017719959,
            0.017524268,
            0.017384537,
            0.016986401,
            0.016817675,
            0.016751431,
            0.016496031
          ],
          "watts": {
            "core": 11.939820488541752,
            "dram": 2.5790372754974538,
            "package-0": 18.919943431206562,
            "psys": 31.4800444068464,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 4.676867,
            "dram": 0.952085,
            "package-0": 8.757728,
            "psys": 14.378259,
            "uncore": 1.553341
          },
          "net_j": {
            "core": 3.679534780354562,
            "dram": 0.7528301900992262,
            "package-0": 7.013604626210564,
            "psys": 10.34594752146848,
            "uncore": 1.5530400064675625
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.3333333333333333,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.861509,
            "dram": 0.054931,
            "package-0": 1.030881,
            "psys": 1.423092,
            "uncore": 0.022461
          },
          "logits": {
            "l2": 714.5644081016375,
            "sha256": "f7f64affdcb433a14d202661deba5ad255d97611c0d28dc97208fa573f8d89b8",
            "sum": 24380.198635727167,
            "top1": 78
          },
          "net_j": {
            "core": 0.7961867649306213,
            "dram": 0.04188041437300608,
            "package-0": 0.9166462096578464,
            "psys": 1.1589878298096186,
            "uncore": 0.022404520592125252
          },
          "npu_busy_delta_us": 21625,
          "npu_utilization": 0.8132337366132795,
          "seconds": 0.026591371,
          "watts": {
            "core": 32.39806627495814,
            "dram": 2.0657453126429624,
            "package-0": 38.76750093103511,
            "psys": 53.517060101940594,
            "uncore": 0.844672506731601
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2522.2051840912736,
          "decode_tokens_per_s": 51.568909467527156,
          "end_to_end_ms": 405.99393200000003,
          "handoff_estimate_ms": 8.573653999999996,
          "handoff_plus_first_decode_ms": 26.591371,
          "output_tokens_per_s_e2e": 39.40945600142615,
          "prefill_tokens_per_s": 8756.004703850813,
          "steady_tpot_ms": 18.017717,
          "steady_tpot_p95_ms": 23.08701755,
          "ttft_ms": 115.120998
        },
        "prefill": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 0.494505,
            "dram": 0.215698,
            "package-0": 2.565728,
            "psys": 4.437183,
            "uncore": 1.53088
          },
          "logits": {
            "l2": 840.9680755108035,
            "sha256": "8fb49fad0852bc3258a6ca67a848f5fb785395020a78359d88430d551db8ad9a",
            "sum": 178291.6503057778,
            "top1": 291
          },
          "net_j": {
            "core": 0.21170796434593514,
            "dram": 0.15919860243738482,
            "package-0": 2.0711757232291075,
            "psys": 3.2938069909826924,
            "uncore": 1.5306354858754372
          },
          "npu_busy_delta_us": 4145,
          "npu_utilization": 0.03600559473954526,
          "seconds": 0.115120998,
          "watts": {
            "core": 4.295523914759668,
            "dram": 1.8736633954476316,
            "package-0": 22.287228607938232,
            "psys": 38.54364605143538,
            "uncore": 13.29800841372136
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 723.4091300558687,
            "sha256": "1c462e5b54f0c0f9e4e380bd8290810c4daf6d05e264abf4482ad7bacb911518",
            "sum": 73007.79722458497,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 3.320853,
            "dram": 0.681456,
            "package-0": 5.161119,
            "psys": 8.517984,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.671640051078006,
            "dram": 0.5517511732888354,
            "package-0": 4.0257826933236105,
            "psys": 5.893152700676169,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 254059,
          "npu_utilization": 0.9613194243141356,
          "seconds": 0.264281563,
          "token_latencies_s": [
            0.022965451,
            0.023312784,
            0.021587818,
            0.020058527,
            0.019119158,
            0.01823842,
            0.017797014,
            0.017611399,
            0.01730654,
            0.019252841,
            0.01700627,
            0.016841375,
            0.01667363,
            0.016487893
          ],
          "watts": {
            "core": 12.565587104538201,
            "dram": 2.578522664481139,
            "package-0": 19.528865129346915,
            "psys": 32.23071599587899,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.8849987629801035,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393081,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU+GPU",
      "key": "npu_gpu_combined",
      "label": "NPU + iGPU (subgraph combination)",
      "mode": "combined",
      "prefill_device": "NPU+GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03892374038696289,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 24.098327,
            "dram": 2.623834,
            "package-0": 38.396387,
            "psys": 63.814412,
            "uncore": 6.060897
          },
          "net_j": {
            "core": 19.08077772474327,
            "dram": 1.621388869565455,
            "package-0": 29.621753234305892,
            "psys": 43.52797069370386,
            "uncore": 6.056558688370768
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.17391304347826086,
          "gpu_freq_mhz_after": 2050,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 69,
          "gross_j": {
            "core": 10.890964,
            "dram": 0.583129,
            "package-0": 12.970182,
            "psys": 19.720348,
            "uncore": 0.358275
          },
          "logits": {
            "l2": 958.485918292106,
            "sha256": "0de2e58c6569eebaed9a2446c17498b4a5555fdcc3ad76d584b7a237d8eacab0",
            "sum": 107240.04813884199,
            "top1": 96630
          },
          "net_j": {
            "core": 9.169017531592287,
            "dram": 0.23910510358575832,
            "package-0": 9.958861374028679,
            "psys": 12.758350409622643,
            "uncore": 0.3567861575434554
          },
          "npu_busy_delta_us": 5730,
          "npu_utilization": 0.008174388658550167,
          "seconds": 0.700969851,
          "watts": {
            "core": 15.536993473346973,
            "dram": 0.8318888453877312,
            "package-0": 18.503195225153842,
            "psys": 28.13294747536867,
            "uncore": 0.5111132803912847
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 501.33568334957124,
          "decode_tokens_per_s": 8.23159884768899,
          "end_to_end_ms": 2042.5436170000003,
          "handoff_estimate_ms": 622.4121865,
          "handoff_plus_first_decode_ms": 700.969851,
          "output_tokens_per_s_e2e": 7.833370052337051,
          "prefill_tokens_per_s": 4575.63311135657,
          "steady_tpot_ms": 78.5576645,
          "steady_tpot_p95_ms": 86.10422714999999,
          "ttft_ms": 220.29738300000002
        },
        "prefill": {
          "gpu_active_fraction": 0.13636363636363635,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 800,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 0.807554,
            "dram": 0.256103,
            "package-0": 2.566644,
            "psys": 4.804187,
            "uncore": 0.042908
          },
          "logits": {
            "l2": 839.9427313198568,
            "sha256": "a824a46e6afb71eec854c6d4700ff61ce067e09421e5019d2f0eb31fb7898035",
            "sum": 177091.2838851288,
            "top1": 291
          },
          "net_j": {
            "core": 0.2663890695609139,
            "dram": 0.1479848491816697,
            "package-0": 1.620260853379613,
            "psys": 2.616204382167784,
            "uncore": 0.04244009386051602
          },
          "npu_busy_delta_us": 180194,
          "npu_utilization": 0.8179579691148668,
          "seconds": 0.220297383,
          "watts": {
            "core": 3.6657448627067892,
            "dram": 1.1625331019025316,
            "package-0": 11.650814753437176,
            "psys": 21.80773522852062,
            "uncore": 0.19477308089492829
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1318.474203054553,
            "sha256": "3478025624cc65876a42b2a858ed37b3779732cb44a02c80b3e2714ea2d2dced",
            "sum": 371578.8326460272,
            "top1": 81826
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2050,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 111,
          "gross_j": {
            "core": 12.399809,
            "dram": 1.784602,
            "package-0": 22.859561,
            "psys": 39.289877,
            "uncore": 5.659714
          },
          "net_j": {
            "core": 9.645371123590069,
            "dram": 1.234298916798027,
            "package-0": 18.0426310068976,
            "psys": 28.153415901913426,
            "uncore": 5.657332436966797
          },
          "npu_busy_delta_us": 73315,
          "npu_utilization": 0.06538530652348538,
          "seconds": 1.121276383,
          "token_latencies_s": [
            0.097667681,
            0.079877752,
            0.07968476,
            0.079762935,
            0.078480599,
            0.078007595,
            0.078859713,
            0.078330556,
            0.078517422,
            0.078050959,
            0.078337537,
            0.078597907,
            0.078286445,
            0.078760557
          ],
          "watts": {
            "core": 11.058655286062507,
            "dram": 1.5915808332868453,
            "package-0": 20.38708863093926,
            "psys": 35.040314409261924,
            "uncore": 5.047563728094681
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 15.588095,
            "dram": 2.195795,
            "package-0": 28.654346,
            "psys": 48.830381,
            "uncore": 6.131087
          },
          "net_j": {
            "core": 12.018391142543287,
            "dram": 1.4826117201371836,
            "package-0": 22.41168802439587,
            "psys": 34.397719991115494,
            "uncore": 6.128000535491503
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.668267,
            "dram": 0.144348,
            "package-0": 3.418204,
            "psys": 4.933154,
            "uncore": 0.355284
          },
          "logits": {
            "l2": 1318.5428560315058,
            "sha256": "ecc3520c68a3ff34e89690bcb65dd6268467d08c74ede766137410ed2d52fabb",
            "sum": 371439.9113267623,
            "top1": 81826
          },
          "net_j": {
            "core": 2.4174995634001264,
            "dram": 0.09424772556308686,
            "package-0": 2.9796647259809674,
            "psys": 3.9192767852409314,
            "uncore": 0.3550671795510612
          },
          "npu_busy_delta_us": 4350,
          "npu_utilization": 0.04261263825609257,
          "seconds": 0.102082391,
          "watts": {
            "core": 26.138366998084912,
            "dram": 1.4140342774690693,
            "package-0": 33.484756445408884,
            "psys": 48.325219968642784,
            "uncore": 3.4803651885465734
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 704.6737194729557,
          "decode_tokens_per_s": 12.094934056625442,
          "end_to_end_ms": 1453.154803,
          "handoff_estimate_ms": 21.9319895,
          "handoff_plus_first_decode_ms": 102.082391,
          "output_tokens_per_s_e2e": 11.010526866764932,
          "prefill_tokens_per_s": 4733.146603186907,
          "steady_tpot_ms": 80.1504015,
          "steady_tpot_p95_ms": 86.93882674999999,
          "ttft_ms": 212.966148
        },
        "prefill": {
          "gpu_active_fraction": 0.09090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1750,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 0.678465,
            "dram": 0.248718,
            "package-0": 2.406855,
            "psys": 4.640552,
            "uncore": 0.058045
          },
          "logits": {
            "l2": 839.9427313198568,
            "sha256": "a824a46e6afb71eec854c6d4700ff61ce067e09421e5019d2f0eb31fb7898035",
            "sum": 177091.2838851288,
            "top1": 291
          },
          "net_j": {
            "core": 0.15530939506800623,
            "dram": 0.14419789270184452,
            "package-0": 1.4919663593670651,
            "psys": 2.52538284244953,
            "uncore": 0.057592665220555735
          },
          "npu_busy_delta_us": 184692,
          "npu_utilization": 0.8672364210672581,
          "seconds": 0.212966148,
          "watts": {
            "core": 3.1857880060825443,
            "dram": 1.1678757508446835,
            "package-0": 11.30158488850538,
            "psys": 21.790092198127187,
            "uncore": 0.27255505414879366
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1316.4160478001859,
            "sha256": "ca8d3e330d15bc8bb96b83f7f7b690e76f28ebabb35d42beaeffef5da51f3929",
            "sum": 369103.7374538556,
            "top1": 89372
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 113,
          "gross_j": {
            "core": 12.241363,
            "dram": 1.802729,
            "package-0": 22.829287,
            "psys": 39.256675,
            "uncore": 5.717758
          },
          "net_j": {
            "core": 9.445582184075155,
            "dram": 1.2441661018722523,
            "package-0": 17.940056939047835,
            "psys": 27.95306036342503,
            "uncore": 5.715340690719886
          },
          "npu_busy_delta_us": 62586,
          "npu_utilization": 0.05499135008714793,
          "seconds": 1.138106264,
          "token_latencies_s": [
            0.097083055,
            0.08147655,
            0.079121465,
            0.079371177,
            0.078459493,
            0.080402221,
            0.079898582,
            0.079365039,
            0.080583146,
            0.081307318,
            0.080788845,
            0.079793174,
            0.079416239,
            0.080987266
          ],
          "watts": {
            "core": 10.755905126975033,
            "dram": 1.583972478689389,
            "package-0": 20.05901181824969,
            "psys": 34.4929785923751,
            "uncore": 5.023922792502968
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 15.326438,
            "dram": 2.207942,
            "package-0": 28.527821,
            "psys": 48.314634999999996,
            "uncore": 6.230819
          },
          "net_j": {
            "core": 11.694591133219905,
            "dram": 1.482343305027267,
            "package-0": 22.176488028297715,
            "psys": 33.63072374082853,
            "uncore": 6.2276788049299805
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9230769230769231,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.770745,
            "dram": 0.163818,
            "package-0": 3.732412,
            "psys": 5.484788,
            "uncore": 0.495909
          },
          "logits": {
            "l2": 1315.6363375493609,
            "sha256": "4a59eb917d5b1c11e3987b53e9dab89eeae07096734a9e6ce2de62e2cd3c22a2",
            "sum": 368620.2270870283,
            "top1": 89372
          },
          "net_j": {
            "core": 2.4429652531692607,
            "dram": 0.09833160553928387,
            "package-0": 3.159194463615266,
            "psys": 4.159542500515758,
            "uncore": 0.4956255925814591
          },
          "npu_busy_delta_us": 4474,
          "npu_utilization": 0.03353004769293299,
          "seconds": 0.133432557,
          "watts": {
            "core": 20.76513455408038,
            "dram": 1.2277213573895611,
            "package-0": 27.972273663315917,
            "psys": 41.105320345468606,
            "uncore": 3.7165517258280523
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 692.6163428473894,
          "decode_tokens_per_s": 11.788574398479327,
          "end_to_end_ms": 1478.451975,
          "handoff_estimate_ms": 53.18247600000001,
          "handoff_plus_first_decode_ms": 133.432557,
          "output_tokens_per_s_e2e": 10.82213035699046,
          "prefill_tokens_per_s": 4892.408558360183,
          "steady_tpot_ms": 80.250081,
          "steady_tpot_p95_ms": 86.65143395,
          "ttft_ms": 206.03348799999998
        },
        "prefill": {
          "gpu_active_fraction": 0.047619047619047616,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2400,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 0.540587,
            "dram": 0.243347,
            "package-0": 2.286555,
            "psys": 4.354787,
            "uncore": 0.078552
          },
          "logits": {
            "l2": 839.9427313198568,
            "sha256": "a824a46e6afb71eec854c6d4700ff61ce067e09421e5019d2f0eb31fb7898035",
            "sum": 177091.2838851288,
            "top1": 291
          },
          "net_j": {
            "core": 0.03446161311039597,
            "dram": 0.14222932249581172,
            "package-0": 1.40144861322729,
            "psys": 2.3084726757152563,
            "uncore": 0.07811439001725
          },
          "npu_busy_delta_us": 183843,
          "npu_utilization": 0.8922966930502095,
          "seconds": 0.206033488,
          "watts": {
            "core": 2.623782207676842,
            "dram": 1.1811041125508686,
            "package-0": 11.097977431707607,
            "psys": 21.136306734757603,
            "uncore": 0.3812584097979257
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1313.8573886044735,
            "sha256": "cde0be62f1d20f004a2b7e7db6ca788a3e41569493213ae305108322f77e0c23",
            "sum": 367252.4857843369,
            "top1": 89372
          },
          "gpu_active_fraction": 0.9910714285714286,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 112,
          "gross_j": {
            "core": 12.015106,
            "dram": 1.800777,
            "package-0": 22.508854,
            "psys": 38.47506,
            "uncore": 5.656358
          },
          "net_j": {
            "core": 9.217164266940248,
            "dram": 1.2417823769921714,
            "package-0": 17.61584495145516,
            "psys": 27.162708564597516,
            "uncore": 5.653938822331272
          },
          "npu_busy_delta_us": 74318,
          "npu_utilization": 0.06524926958491928,
          "seconds": 1.13898593,
          "token_latencies_s": [
            0.096948186,
            0.080219025,
            0.080689727,
            0.080281137,
            0.080410467,
            0.079389672,
            0.079738538,
            0.079822903,
            0.079175359,
            0.079584611,
            0.079686347,
            0.081107029,
            0.080941774,
            0.080919268
          ],
          "watts": {
            "core": 10.548950328122139,
            "dram": 1.5810353337727359,
            "package-0": 19.762187931505,
            "psys": 33.780101216878066,
            "uncore": 4.966135095277252
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.9877989795058966,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 392786,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04277300927788019,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 5.998336,
            "dram": 0.998837,
            "package-0": 11.695039999999999,
            "psys": 18.551467,
            "uncore": 3.541006
          },
          "net_j": {
            "core": 4.895510088230332,
            "dram": 0.7785058374497378,
            "package-0": 9.766430443375091,
            "psys": 14.092634230682913,
            "uncore": 3.5400524662695707
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.713984,
            "dram": 0.072143,
            "package-0": 2.084529,
            "psys": 2.777031,
            "uncore": 0.187622
          },
          "logits": {
            "l2": 712.6588257239783,
            "sha256": "990e8b5de951b6913e147a016df0e6896d435e87a38f418db97e9b4ce13bdab3",
            "sum": 24349.479562914465,
            "top1": 78
          },
          "net_j": {
            "core": 1.585145629701802,
            "dram": 0.04640266545950007,
            "package-0": 1.8592179064128376,
            "psys": 2.2561248975044728,
            "uncore": 0.18751060278803594
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.052447515,
          "watts": {
            "core": 32.67998493350924,
            "dram": 1.3755275154599793,
            "package-0": 39.745047977964255,
            "psys": 52.948762205416216,
            "uncore": 3.5773286875460166
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2280.9370616024767,
          "decode_tokens_per_s": 56.051410832320784,
          "end_to_end_ms": 448.938297,
          "handoff_estimate_ms": 37.251965999999996,
          "handoff_plus_first_decode_ms": 52.447515,
          "output_tokens_per_s_e2e": 35.6396415875387,
          "prefill_tokens_per_s": 5559.022744758106,
          "steady_tpot_ms": 15.195549,
          "steady_tpot_p95_ms": 16.1355517,
          "ttft_ms": 181.326835
        },
        "prefill": {
          "gpu_active_fraction": 0.6111111111111112,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 18,
          "gross_j": {
            "core": 1.033628,
            "dram": 0.286864,
            "package-0": 3.569755,
            "psys": 6.132004,
            "uncore": 1.851008
          },
          "logits": {
            "l2": 838.9345161742121,
            "sha256": "e1f6e0509b78af0035fc75a4b5671053a3706868fabf97ca68685fc192833e98",
            "sum": 177352.29018324683,
            "top1": 291
          },
          "net_j": {
            "core": 0.5881950006914488,
            "dram": 0.19787192107938706,
            "package-0": 2.790786778330223,
            "psys": 4.331074920753833,
            "uncore": 1.8506228663311644
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.181326835,
          "watts": {
            "core": 5.700358692082173,
            "dram": 1.582027282393144,
            "package-0": 19.686854402990047,
            "psys": 33.81741042355921,
            "uncore": 10.2081305285012
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 720.1119452755735,
            "sha256": "9d31e7729784e25aa0cb77989028372b1f68cbebfb778541584dec15714a8c90",
            "sum": 72014.2167466206,
            "top1": 42
          },
          "gpu_active_fraction": 0.9545454545454546,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.250724,
            "dram": 0.63983,
            "package-0": 6.040756,
            "psys": 9.642432,
            "uncore": 1.502376
          },
          "net_j": {
            "core": 2.722169457837081,
            "dram": 0.5342312509108507,
            "package-0": 5.11642575863203,
            "psys": 7.5054344124246075,
            "uncore": 1.5019189971503706
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.215163947,
          "token_latencies_s": [
            0.01732816,
            0.015319979,
            0.015055705,
            0.015043114,
            0.015227711,
            0.015446988,
            0.015493378,
            0.015282708,
            0.015270521,
            0.015133025,
            0.015146154,
            0.015163387,
            0.015136484,
            0.015087242
          ],
          "watts": {
            "core": 15.108125898062282,
            "dram": 2.9736859214615543,
            "package-0": 28.07513100696187,
            "psys": 44.81434800970629,
            "uncore": 6.982470906243415
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.037643,
            "dram": 1.026122,
            "package-0": 11.773589,
            "psys": 18.811109,
            "uncore": 3.5179970000000003
          },
          "net_j": {
            "core": 4.927513753510475,
            "dram": 0.8043317202743555,
            "package-0": 9.832207453720489,
            "psys": 14.32274813578901,
            "uncore": 3.51703715160474
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.88513,
            "dram": 0.055054,
            "package-0": 1.073178,
            "psys": 1.45898,
            "uncore": 0.066528
          },
          "logits": {
            "l2": 712.6588257239783,
            "sha256": "990e8b5de951b6913e147a016df0e6896d435e87a38f418db97e9b4ce13bdab3",
            "sum": 24349.479562914465,
            "top1": 78
          },
          "net_j": {
            "core": 0.8200700819638291,
            "dram": 0.04205582211593866,
            "package-0": 0.9594019467365738,
            "psys": 1.1959364031704132,
            "uncore": 0.06647174739867429
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.026484587,
          "watts": {
            "core": 33.420570235813,
            "dram": 2.0787184636860676,
            "package-0": 40.52085086318318,
            "psys": 55.087889420363624,
            "uncore": 2.511951573947519
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2265.9311990972883,
          "decode_tokens_per_s": 61.55369402594316,
          "end_to_end_ms": 451.911338,
          "handoff_estimate_ms": 11.253072999999999,
          "handoff_plus_first_decode_ms": 26.484587,
          "output_tokens_per_s_e2e": 35.40517498589513,
          "prefill_tokens_per_s": 4840.9949746918255,
          "steady_tpot_ms": 15.231514,
          "steady_tpot_p95_ms": 16.6062088,
          "ttft_ms": 208.221658
        },
        "prefill": {
          "gpu_active_fraction": 0.5714285714285714,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 1.734492,
            "dram": 0.334961,
            "package-0": 4.465076,
            "psys": 7.540264,
            "uncore": 1.931453
          },
          "logits": {
            "l2": 838.9345161742121,
            "sha256": "e1f6e0509b78af0035fc75a4b5671053a3706868fabf97ca68685fc192833e98",
            "sum": 177352.29018324683,
            "top1": 291
          },
          "net_j": {
            "core": 1.2229913297106885,
            "dram": 0.23276940645171534,
            "package-0": 3.570569378046374,
            "psys": 5.472216880437966,
            "uncore": 1.9310107423988205
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.208221658,
          "watts": {
            "core": 8.330026840915847,
            "dram": 1.6086751167834807,
            "package-0": 21.443859600810594,
            "psys": 36.21267870223183,
            "uncore": 9.275946693306995
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 720.1119452755735,
            "sha256": "9d31e7729784e25aa0cb77989028372b1f68cbebfb778541584dec15714a8c90",
            "sum": 72014.2167466206,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.418021,
            "dram": 0.636107,
            "package-0": 6.235335,
            "psys": 9.811865,
            "uncore": 1.520016
          },
          "net_j": {
            "core": 2.8844523418359573,
            "dram": 0.5295064917067015,
            "package-0": 5.302236128937542,
            "psys": 7.654594852180631,
            "uncore": 1.519554661807245
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.217205093,
          "token_latencies_s": [
            0.018152538,
            0.01577357,
            0.015673576,
            0.015386031,
            0.015580588,
            0.01542957,
            0.01522996,
            0.015233068,
            0.015164214,
            0.015094769,
            0.015152342,
            0.015167017,
            0.015092245,
            0.015043933
          ],
          "watts": {
            "core": 15.736375942160805,
            "dram": 2.9286007579941966,
            "package-0": 28.707130730125193,
            "psys": 45.17327316998041,
            "uncore": 6.998067950460076
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 5.887314,
            "dram": 1.042111,
            "package-0": 11.709137,
            "psys": 18.89546,
            "uncore": 3.5533360000000003
          },
          "net_j": {
            "core": 4.736853507198252,
            "dram": 0.8122630293076882,
            "package-0": 9.697224623570218,
            "psys": 14.24403597163487,
            "uncore": 3.5523412800957024
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.804808,
            "dram": 0.053283,
            "package-0": 0.978452,
            "psys": 1.334348,
            "uncore": 0.062561
          },
          "logits": {
            "l2": 712.6588257239783,
            "sha256": "990e8b5de951b6913e147a016df0e6896d435e87a38f418db97e9b4ce13bdab3",
            "sum": 24349.479562914465,
            "top1": 78
          },
          "net_j": {
            "core": 0.746232909127959,
            "dram": 0.04158041146758662,
            "package-0": 0.8760165395793122,
            "psys": 1.0975231921507893,
            "uncore": 0.06251035435927524
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.023844744,
          "watts": {
            "core": 33.752008409064906,
            "dram": 2.234580501262668,
            "package-0": 41.03428411728807,
            "psys": 55.95983752226486,
            "uncore": 2.6236809252387028
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2186.495329817863,
          "decode_tokens_per_s": 61.19568611046857,
          "end_to_end_ms": 468.329379,
          "handoff_estimate_ms": 8.4692085,
          "handoff_plus_first_decode_ms": 23.844744000000002,
          "output_tokens_per_s_e2e": 34.16398952840411,
          "prefill_tokens_per_s": 4515.844525960018,
          "steady_tpot_ms": 15.3755355,
          "steady_tpot_p95_ms": 17.291089500000002,
          "ttft_ms": 223.21406199999998
        },
        "prefill": {
          "gpu_active_fraction": 0.5454545454545454,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 1.633907,
            "dram": 0.345092,
            "package-0": 4.422962,
            "psys": 7.600811,
            "uncore": 1.944514
          },
          "logits": {
            "l2": 838.9345161742121,
            "sha256": "e1f6e0509b78af0035fc75a4b5671053a3706868fabf97ca68685fc192833e98",
            "sum": 177352.29018324683,
            "top1": 291
          },
          "net_j": {
            "core": 1.0855771892221031,
            "dram": 0.23554239350245876,
            "package-0": 3.4640489911900945,
            "psys": 5.3838600718461125,
            "uncore": 1.9440398989071652
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.223214062,
          "watts": {
            "core": 7.319910696307296,
            "dram": 1.5460137094767803,
            "package-0": 19.81488961927497,
            "psys": 34.05166740794314,
            "uncore": 8.711431450945058
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 720.1119452755735,
            "sha256": "9d31e7729784e25aa0cb77989028372b1f68cbebfb778541584dec15714a8c90",
            "sum": 72014.2167466206,
            "top1": 42
          },
          "gpu_active_fraction": 0.9545454545454546,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.448599,
            "dram": 0.643736,
            "package-0": 6.307723,
            "psys": 9.960301,
            "uncore": 1.546261
          },
          "net_j": {
            "core": 2.90504340884819,
            "dram": 0.5351402243376429,
            "package-0": 5.357159092800812,
            "psys": 7.762652707637969,
            "uncore": 1.5457910268292618
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.221270573,
          "token_latencies_s": [
            0.019060071,
            0.016338561,
            0.016201103,
            0.016299304,
            0.016259874,
            0.01551903,
            0.015116966,
            0.01501455,
            0.015250364,
            0.015247183,
            0.015074194,
            0.015232958,
            0.015500707,
            0.015121262
          ],
          "watts": {
            "core": 15.585438918712432,
            "dram": 2.909270723495618,
            "package-0": 28.506831769265588,
            "psys": 45.014123952216636,
            "uncore": 6.988100491790203
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 0.9142003450542688,
    "compile_target": "AUTO:GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395961,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03955871798098087,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 7.614482,
            "dram": 1.349363,
            "package-0": 16.965653,
            "psys": 27.241691,
            "uncore": 6.301742
          },
          "net_j": {
            "core": 5.008566568409632,
            "dram": 1.0081887576951474,
            "package-0": 13.00904908862839,
            "psys": 20.224601601738687,
            "uncore": 6.300392615720058
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.621517,
            "dram": 0.070496,
            "package-0": 1.858332,
            "psys": 2.546014,
            "uncore": 0.09967
          },
          "logits": {
            "l2": 752.85195897198,
            "sha256": "4970ddf30ee86b82fca779e34e85f3a0321c66b97a1a267f9f290f6c9784648a",
            "sum": -146160.83817428164,
            "top1": 522
          },
          "net_j": {
            "core": 1.4451380894002102,
            "dram": 0.0474039454947186,
            "package-0": 1.5905330226798906,
            "psys": 2.0710689696882034,
            "uncore": 0.0995786682068069
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.046763065,
          "watts": {
            "core": 34.67516511161106,
            "dram": 1.5075145309658382,
            "package-0": 39.7393113560884,
            "psys": 54.444977034760235,
            "uncore": 2.131382962173245
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2964.23971744647,
          "decode_tokens_per_s": 49.79433908873345,
          "end_to_end_ms": 690.902287,
          "handoff_estimate_ms": 30.324757,
          "handoff_plus_first_decode_ms": 46.763065,
          "output_tokens_per_s_e2e": 23.158122792550547,
          "prefill_tokens_per_s": 5214.759488786759,
          "steady_tpot_ms": 16.438308,
          "steady_tpot_p95_ms": 26.155030949999997,
          "ttft_ms": 389.663225
        },
        "prefill": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 39,
          "gross_j": {
            "core": 2.511834,
            "dram": 0.614562,
            "package-0": 8.523781,
            "psys": 14.199976,
            "uncore": 4.516529
          },
          "logits": {
            "l2": 658.700794245309,
            "sha256": "d8b9f589d3ed262681597fb65c434b4174e3993033527e14721527676e3f0747",
            "sum": -95245.78444925009,
            "top1": 220
          },
          "net_j": {
            "core": 1.042119063087699,
            "dram": 0.42214252470675884,
            "package-0": 6.292288835099008,
            "psys": 10.242395114123134,
            "uncore": 4.515767958330027
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.389663225,
          "watts": {
            "core": 6.446166429998622,
            "dram": 1.5771619197577602,
            "package-0": 21.874738115201914,
            "psys": 36.44166318235445,
            "uncore": 11.59085258815481
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 726.9945112251366,
            "sha256": "82a7d15255d9e580be3390c3a34d6d3e2feb60a349b88eded778069599535d54",
            "sum": -57824.15603592619,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 25,
          "gross_j": {
            "core": 3.481131,
            "dram": 0.664305,
            "package-0": 6.58354,
            "psys": 10.495701,
            "uncore": 1.685543
          },
          "net_j": {
            "core": 2.5213094159217233,
            "dram": 0.5386422874936699,
            "package-0": 5.126227230849491,
            "psys": 7.91113751792735,
            "uncore": 1.6850459891832237
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.254475997,
          "token_latencies_s": [
            0.026521564,
            0.025957667,
            0.021161474,
            0.016808544,
            0.016487786,
            0.016380679,
            0.016314393,
            0.016352132,
            0.016308652,
            0.016494606,
            0.016401723,
            0.016397908,
            0.016474893,
            0.016375879
          ],
          "watts": {
            "core": 13.679604524744233,
            "dram": 2.6104819622732434,
            "package-0": 25.870966525774136,
            "psys": 41.244365377218664,
            "uncore": 6.623583441545569
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 7.980022,
            "dram": 1.427792,
            "package-0": 17.574966,
            "psys": 28.345996,
            "uncore": 6.388472999999999
          },
          "net_j": {
            "core": 5.31761053903866,
            "dram": 1.0792211285566065,
            "package-0": 13.532583245586757,
            "psys": 21.176776688412616,
            "uncore": 6.3870943611813304
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.971921,
            "dram": 0.057922,
            "package-0": 1.173642,
            "psys": 1.590877,
            "uncore": 0.07434
          },
          "logits": {
            "l2": 752.85195897198,
            "sha256": "4970ddf30ee86b82fca779e34e85f3a0321c66b97a1a267f9f290f6c9784648a",
            "sum": -146160.83817428164,
            "top1": 522
          },
          "net_j": {
            "core": 0.8622817690720452,
            "dram": 0.043567703968258424,
            "package-0": 1.0071749670830579,
            "psys": 1.295645514107709,
            "uncore": 0.07428322706988659
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.029068478,
          "watts": {
            "core": 33.43556549469154,
            "dram": 1.9926051855896962,
            "package-0": 40.375075709158224,
            "psys": 54.72859638540415,
            "uncore": 2.557409438498982
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2901.3389312250233,
          "decode_tokens_per_s": 56.753306339796076,
          "end_to_end_ms": 705.8809909999999,
          "handoff_estimate_ms": 12.520836999999997,
          "handoff_plus_first_decode_ms": 29.068478,
          "output_tokens_per_s_e2e": 22.666710400195495,
          "prefill_tokens_per_s": 4601.665926712596,
          "steady_tpot_ms": 16.547641000000002,
          "steady_tpot_p95_ms": 18.1297297,
          "ttft_ms": 441.579209
        },
        "prefill": {
          "gpu_active_fraction": 0.6136363636363636,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 44,
          "gross_j": {
            "core": 3.536735,
            "dram": 0.712462,
            "package-0": 9.944432,
            "psys": 16.56496,
            "uncore": 4.677235
          },
          "logits": {
            "l2": 658.700794245309,
            "sha256": "d8b9f589d3ed262681597fb65c434b4174e3993033527e14721527676e3f0747",
            "sum": -95245.78444925009,
            "top1": 220
          },
          "net_j": {
            "core": 1.8712055949676967,
            "dram": 0.49440590888135655,
            "package-0": 7.415631533734093,
            "psys": 12.080098907093362,
            "uncore": 4.676372562497352
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.441579209,
          "watts": {
            "core": 8.009287864818836,
            "dram": 1.613440998758617,
            "package-0": 22.520154475841732,
            "psys": 37.512998036100925,
            "uncore": 10.592063450161215
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 726.9945112251366,
            "sha256": "82a7d15255d9e580be3390c3a34d6d3e2feb60a349b88eded778069599535d54",
            "sum": -57824.15603592619,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 23,
          "gross_j": {
            "core": 3.471366,
            "dram": 0.657408,
            "package-0": 6.456892,
            "psys": 10.190159,
            "uncore": 1.636898
          },
          "net_j": {
            "core": 2.5841231749989184,
            "dram": 0.5412475157069916,
            "package-0": 5.109776744769607,
            "psys": 7.801032267211546,
            "uncore": 1.6364385716140921
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.235233304,
          "token_latencies_s": [
            0.020142235,
            0.017046073,
            0.016604627,
            0.016623763,
            0.016571168,
            0.016388021,
            0.016310621,
            0.016478148,
            0.016615487,
            0.016391979,
            0.016413113,
            0.016524114,
            0.016422428,
            0.016665236
          ],
          "watts": {
            "core": 14.757119595616444,
            "dram": 2.7947063142045567,
            "package-0": 27.44888538401858,
            "psys": 43.31937198824534,
            "uncore": 6.958615009718182
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.231913,
            "dram": 1.4384130000000002,
            "package-0": 17.964798000000002,
            "psys": 28.874437,
            "uncore": 6.444440999999999
          },
          "net_j": {
            "core": 5.523284780440212,
            "dram": 1.0837912922917516,
            "package-0": 13.852243587049882,
            "psys": 21.580767323350216,
            "uncore": 6.443038429411801
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 1900,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 4,
          "gross_j": {
            "core": 1.338681,
            "dram": 0.0625,
            "package-0": 1.614803,
            "psys": 2.114984,
            "uncore": 0.094055
          },
          "logits": {
            "l2": 752.85195897198,
            "sha256": "4970ddf30ee86b82fca779e34e85f3a0321c66b97a1a267f9f290f6c9784648a",
            "sum": -146160.83817428164,
            "top1": 522
          },
          "net_j": {
            "core": 1.197080614412732,
            "dram": 0.043961254008029055,
            "package-0": 1.3998088245656908,
            "psys": 1.7336889846480503,
            "uncore": 0.0939816770778403
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.037542289,
          "watts": {
            "core": 35.657948293989215,
            "dram": 1.6647892727052418,
            "package-0": 43.012907390915885,
            "psys": 56.33604280229158,
            "uncore": 2.5053080807086645
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2851.833989930976,
          "decode_tokens_per_s": 54.51941527970728,
          "end_to_end_ms": 718.1343680000001,
          "handoff_estimate_ms": 20.8518715,
          "handoff_plus_first_decode_ms": 37.542289,
          "output_tokens_per_s_e2e": 22.27995304633575,
          "prefill_tokens_per_s": 4586.876138344373,
          "steady_tpot_ms": 16.6904175,
          "steady_tpot_p95_ms": 18.34303435,
          "ttft_ms": 443.00302400000004
        },
        "prefill": {
          "gpu_active_fraction": 0.6363636363636364,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 44,
          "gross_j": {
            "core": 3.416251,
            "dram": 0.715453,
            "package-0": 9.854345,
            "psys": 16.48769,
            "uncore": 4.696338
          },
          "logits": {
            "l2": 658.700794245309,
            "sha256": "d8b9f589d3ed262681597fb65c434b4174e3993033527e14721527676e3f0747",
            "sum": -95245.78444925009,
            "top1": 220
          },
          "net_j": {
            "core": 1.745351310992605,
            "dram": 0.49669381524142,
            "package-0": 7.317390743966612,
            "psys": 11.988368053034002,
            "uncore": 4.695472781679261
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.443003024,
          "watts": {
            "core": 7.711574898865701,
            "dram": 1.615007034353788,
            "package-0": 22.24441926157145,
            "psys": 37.21800779400549,
            "uncore": 10.601142081594459
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 726.9945112251366,
            "sha256": "82a7d15255d9e580be3390c3a34d6d3e2feb60a349b88eded778069599535d54",
            "sum": -57824.15603592619,
            "top1": 42
          },
          "gpu_active_fraction": 0.9583333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 24,
          "gross_j": {
            "core": 3.476981,
            "dram": 0.66046,
            "package-0": 6.49565,
            "psys": 10.271763,
            "uncore": 1.654048
          },
          "net_j": {
            "core": 2.580852855034875,
            "dram": 0.5431362230423027,
            "package-0": 5.135044018517579,
            "psys": 7.858710285668163,
            "uncore": 1.6535839706546993
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.237589055,
          "token_latencies_s": [
            0.020231909,
            0.017325948,
            0.017130183,
            0.017158318,
            0.01676068,
            0.016601071,
            0.016842128,
            0.01662665,
            0.016534849,
            0.016754185,
            0.016387876,
            0.01641993,
            0.016365636,
            0.016426358
          ],
          "watts": {
            "core": 14.63443254993375,
            "dram": 2.7798418576141906,
            "package-0": 27.339853681391176,
            "psys": 43.23331729233066,
            "uncore": 6.961802175609479
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.1983783105388284,
    "compile_target": "NPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396029,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 2048,
      "NPUW_LLM_MIN_RESPONSE_LEN": 2048,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03935159929096699,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 8.949988000000001,
            "dram": 1.463071,
            "package-0": 16.844744,
            "psys": 27.576833999999998,
            "uncore": 4.024343
          },
          "net_j": {
            "core": 6.162054949215223,
            "dram": 1.0980664665959126,
            "package-0": 12.611779768062025,
            "psys": 20.069615915879385,
            "uncore": 4.023731913936043
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 16,
          "gross_j": {
            "core": 2.839958,
            "dram": 0.134277,
            "package-0": 3.244681,
            "psys": 4.804675,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1443.2633782892876,
            "sha256": "8dfbc68a4d27bc8c6dd2a02d6083dc30ff3beb0b60761b166fbe665205c7e7e4",
            "sum": -388910.4892301038,
            "top1": 8
          },
          "net_j": {
            "core": 2.246151140195466,
            "dram": 0.05653402990461359,
            "package-0": 2.34309435769363,
            "psys": 3.2056991583295424,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 24839,
          "npu_utilization": 0.15777296401243726,
          "seconds": 0.157435085,
          "watts": {
            "core": 18.038914261074652,
            "dram": 0.8529039127460057,
            "package-0": 20.609643650905387,
            "psys": 30.518451462074033,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2770.711448918199,
          "decode_tokens_per_s": 35.188442366426685,
          "end_to_end_ms": 739.160334,
          "handoff_estimate_ms": 139.6647425,
          "handoff_plus_first_decode_ms": 157.43508500000002,
          "output_tokens_per_s_e2e": 21.64618319467343,
          "prefill_tokens_per_s": 6494.419532585504,
          "steady_tpot_ms": 17.7703425,
          "steady_tpot_p95_ms": 24.969379049999997,
          "ttft_ms": 312.884006
        },
        "prefill": {
          "gpu_active_fraction": 0.967741935483871,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 31,
          "gross_j": {
            "core": 2.344964,
            "dram": 0.628844,
            "package-0": 7.914835,
            "psys": 13.506679,
            "uncore": 4.024343
          },
          "logits": {
            "l2": 1214.4736652113506,
            "sha256": "6309e0e955846c137342cf9ddc36ea672b062c49bbf09e8acb3ff6ad93269bdf",
            "sum": 281703.18898873776,
            "top1": 89372
          },
          "net_j": {
            "core": 1.1648416080571682,
            "dram": 0.4743388468330124,
            "package-0": 6.123035920392642,
            "psys": 10.328899619761636,
            "uncore": 4.023731913936043
          },
          "npu_busy_delta_us": 5856,
          "npu_utilization": 0.018716201172647988,
          "seconds": 0.312884006,
          "watts": {
            "core": 7.494675199217438,
            "dram": 2.009831080978936,
            "package-0": 25.29638731357844,
            "psys": 43.16832673128073,
            "uncore": 12.862092413889638
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1281.626957611578,
            "sha256": "20ba5eabdad4a4b070d6d6d3912744bc0c1fa18ee32a52ccd4fb0399db15ebad",
            "sum": -343781.3722482808,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 27,
          "gross_j": {
            "core": 3.765066,
            "dram": 0.69995,
            "package-0": 5.685228,
            "psys": 9.26548,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.751062200962589,
            "dram": 0.5671935898582865,
            "package-0": 4.145649489975752,
            "psys": 6.535017137788207,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 256774,
          "npu_utilization": 0.9551138699354994,
          "seconds": 0.268841243,
          "token_latencies_s": [
            0.02553388,
            0.024665417,
            0.022123729,
            0.019989949,
            0.0201231,
            0.018341175,
            0.017858462,
            0.017682223,
            0.017409616,
            0.017279776,
            0.017121826,
            0.016946542,
            0.017021695,
            0.016718193
          ],
          "watts": {
            "core": 14.004793155937016,
            "dram": 2.6035811774609297,
            "package-0": 21.147157097469602,
            "psys": 34.464503647604396,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 7.15196,
            "dram": 1.392758,
            "package-0": 14.817345000000001,
            "psys": 24.471799,
            "uncore": 3.953237
          },
          "net_j": {
            "core": 4.77961377566322,
            "dram": 1.0821633368881125,
            "package-0": 11.215373026331571,
            "psys": 18.083654036431717,
            "uncore": 3.95262917805632
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.615658,
            "dram": 0.070984,
            "package-0": 1.806148,
            "psys": 2.437432,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1215.1658719840748,
            "sha256": "67f1824ed2e475f93e13a8b3a0737153bdb4a3bba006f63c1e57e63dd98c2618",
            "sum": -177402.685341008,
            "top1": 8
          },
          "net_j": {
            "core": 1.4294050486398115,
            "dram": 0.046599206546217756,
            "package-0": 1.5233571055132642,
            "psys": 1.9358986061911159,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 23715,
          "npu_utilization": 0.48024588589357753,
          "seconds": 0.049380954,
          "watts": {
            "core": 32.71824193594964,
            "dram": 1.4374772913459712,
            "package-0": 36.5758020794819,
            "psys": 49.359759230248976,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3256.083763568839,
          "decode_tokens_per_s": 47.20488633261632,
          "end_to_end_ms": 628.976448,
          "handoff_estimate_ms": 31.236676999999997,
          "handoff_plus_first_decode_ms": 49.380953999999996,
          "output_tokens_per_s_e2e": 25.438154402881555,
          "prefill_tokens_per_s": 6529.295809598845,
          "steady_tpot_ms": 18.144277,
          "steady_tpot_p95_ms": 23.3607074,
          "ttft_ms": 311.212734
        },
        "prefill": {
          "gpu_active_fraction": 0.967741935483871,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 31,
          "gross_j": {
            "core": 2.538934,
            "dram": 0.623411,
            "package-0": 8.038187,
            "psys": 13.643459,
            "uncore": 3.953237
          },
          "logits": {
            "l2": 1294.8704681928841,
            "sha256": "f67be6ad2198c7836d5609b7d08a700e9459eafcb8c610546c38ffc62933f614",
            "sum": 344172.9917854555,
            "top1": 89372
          },
          "net_j": {
            "core": 1.3651152396663435,
            "dram": 0.46973113713674153,
            "package-0": 6.255958827099116,
            "psys": 10.482653749466333,
            "uncore": 3.95262917805632
          },
          "npu_busy_delta_us": 4171,
          "npu_utilization": 0.013402407884762197,
          "seconds": 0.311212734,
          "watts": {
            "core": 8.158194452287418,
            "dram": 2.0031667470264893,
            "package-0": 25.82859286214169,
            "psys": 43.839655352920104,
            "uncore": 12.702683946088145
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1228.9703970899452,
            "sha256": "063abfdbf40b7782053c5f3c313384a0d3cb22f1fd987cddfffb471ee7aa5ff4",
            "sum": -187181.752090808,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 27,
          "gross_j": {
            "core": 2.997368,
            "dram": 0.698363,
            "package-0": 4.97301,
            "psys": 8.390908,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.9850934873570654,
            "dram": 0.5658329932051532,
            "package-0": 3.4360570937191905,
            "psys": 5.665101680774267,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 257207,
          "npu_utilization": 0.9583588752124019,
          "seconds": 0.26838276,
          "token_latencies_s": [
            0.023516723,
            0.023276699,
            0.022803482,
            0.020529341,
            0.019389887,
            0.018663185,
            0.018351903,
            0.017936651,
            0.01785845,
            0.017481147,
            0.017311715,
            0.017215874,
            0.017055887,
            0.016967341
          ],
          "watts": {
            "core": 11.168258348636105,
            "dram": 2.6021157245718762,
            "package-0": 18.52954340286239,
            "psys": 31.264705676325857,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.682905,
            "dram": 1.384519,
            "package-0": 14.293115,
            "psys": 23.554443,
            "uncore": 3.963246
          },
          "net_j": {
            "core": 4.369907773228524,
            "dram": 1.0816944851555632,
            "package-0": 10.781253577524536,
            "psys": 17.32611029148973,
            "uncore": 3.962576104690427
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.25,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 4,
          "gross_j": {
            "core": 1.210813,
            "dram": 0.063293,
            "package-0": 1.403256,
            "psys": 1.87884,
            "uncore": 0.016784
          },
          "logits": {
            "l2": 1332.6647311881006,
            "sha256": "ce35f508c161565adce06026e2f517cd92518ab10af6c84aa72fcae548a07a08",
            "sum": -237999.51964376122,
            "top1": 151643
          },
          "net_j": {
            "core": 1.0691443493855584,
            "dram": 0.04474531654649845,
            "package-0": 1.188158176661292,
            "psys": 1.4973611637179292,
            "uncore": 0.016710641729128135
          },
          "npu_busy_delta_us": 22856,
          "npu_utilization": 0.6085134157826059,
          "seconds": 0.037560388,
          "watts": {
            "core": 32.23643483129088,
            "dram": 1.6850997385863,
            "package-0": 37.359997452635476,
            "psys": 50.0218474846426,
            "uncore": 0.4468537438963623
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3339.631337737878,
          "decode_tokens_per_s": 48.73189457829246,
          "end_to_end_ms": 613.2413409999999,
          "handoff_estimate_ms": 19.081272,
          "handoff_plus_first_decode_ms": 37.560388,
          "output_tokens_per_s_e2e": 26.09086982607717,
          "prefill_tokens_per_s": 6652.812838734416,
          "steady_tpot_ms": 18.479116,
          "steady_tpot_p95_ms": 23.452067549999995,
          "ttft_ms": 305.434716
        },
        "prefill": {
          "gpu_active_fraction": 0.967741935483871,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 31,
          "gross_j": {
            "core": 2.372369,
            "dram": 0.624938,
            "package-0": 7.854838,
            "psys": 13.233853,
            "uncore": 3.946462
          },
          "logits": {
            "l2": 1317.0179500179788,
            "sha256": "f133d0b67d8a6a6da60f1d83f451f8880173f1bcb958e32ccb1b50ebc8687a23",
            "sum": 359899.42827031016,
            "top1": 89372
          },
          "net_j": {
            "core": 1.2203435186198506,
            "dram": 0.47411137839858986,
            "package-0": 6.105698918058769,
            "psys": 10.131731690182198,
            "uncore": 3.9458654629612986
          },
          "npu_busy_delta_us": 4197,
          "npu_utilization": 0.01374107061228757,
          "seconds": 0.305434716,
          "watts": {
            "core": 7.767188455421027,
            "dram": 2.0460608020733306,
            "package-0": 25.7169129392613,
            "psys": 43.327926744253915,
            "uncore": 12.92080367183932
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1286.3926898566062,
            "sha256": "ec3d7969f0d2cefd54957f9226c67c9e51f1f296cab99825f211f20021ce44f2",
            "sum": -189694.47188457847,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 27,
          "gross_j": {
            "core": 3.099723,
            "dram": 0.696288,
            "package-0": 5.035021,
            "psys": 8.44175,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.080419905223115,
            "dram": 0.562837790210475,
            "package-0": 3.487396482804475,
            "psys": 5.6970174375896026,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 258158,
          "npu_utilization": 0.9552695455293241,
          "seconds": 0.270246237,
          "token_latencies_s": [
            0.023975885,
            0.023170012,
            0.022205197,
            0.020709119,
            0.019696798,
            0.018954403,
            0.018650373,
            0.018307859,
            0.018001305,
            0.017715389,
            0.017528287,
            0.017228937,
            0.017063598,
            0.017014403
          ],
          "watts": {
            "core": 11.46999504751661,
            "dram": 2.5764947098967377,
            "package-0": 18.631234447123866,
            "psys": 31.237252713346756,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.040728172287345,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 395874,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.0431777061894536,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 7.671978,
            "dram": 1.331906,
            "package-0": 17.13582,
            "psys": 26.943474,
            "uncore": 6.3591750000000005
          },
          "net_j": {
            "core": 5.215483435339661,
            "dram": 1.0102943844946957,
            "package-0": 13.406084228464792,
            "psys": 20.328738250097036,
            "uncore": 6.3579029882113085
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 2100,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2100,
          "gpu_sample_count": 4,
          "gross_j": {
            "core": 1.408565,
            "dram": 0.067627,
            "package-0": 1.680172,
            "psys": 2.233698,
            "uncore": 0.109131
          },
          "logits": {
            "l2": 752.85195897198,
            "sha256": "4970ddf30ee86b82fca779e34e85f3a0321c66b97a1a267f9f290f6c9784648a",
            "sum": -146160.83817428164,
            "top1": 522
          },
          "net_j": {
            "core": 1.2604030021787365,
            "dram": 0.048229188109307444,
            "package-0": 1.455215221667988,
            "psys": 1.8347341765645226,
            "uncore": 0.10905427937111035
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.039281959,
          "watts": {
            "core": 35.85780943358757,
            "dram": 1.721579109636564,
            "package-0": 42.77210309190537,
            "psys": 56.863202774586675,
            "uncore": 2.7781455604085328
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3144.545130998449,
          "decode_tokens_per_s": 55.71604541802646,
          "end_to_end_ms": 651.286566,
          "handoff_estimate_ms": 22.8349735,
          "handoff_plus_first_decode_ms": 39.281959,
          "output_tokens_per_s_e2e": 24.566758835925384,
          "prefill_tokens_per_s": 5318.476497280693,
          "steady_tpot_ms": 16.446985499999997,
          "steady_tpot_p95_ms": 16.7151854,
          "ttft_ms": 382.0643
        },
        "prefill": {
          "gpu_active_fraction": 0.6578947368421053,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 38,
          "gross_j": {
            "core": 2.738824,
            "dram": 0.60858,
            "package-0": 8.897682,
            "psys": 14.468713,
            "uncore": 4.599902
          },
          "logits": {
            "l2": 658.700794245309,
            "sha256": "d8b9f589d3ed262681597fb65c434b4174e3993033527e14721527676e3f0747",
            "sum": -95245.78444925009,
            "top1": 220
          },
          "net_j": {
            "core": 1.2977703604861806,
            "dram": 0.4199129474848967,
            "package-0": 6.70970675050505,
            "psys": 10.58830993949583,
            "uncore": 4.599155799603724
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.3820643,
          "watts": {
            "core": 7.16848970186432,
            "dram": 1.5928732414936437,
            "package-0": 23.28844123881765,
            "psys": 37.869837616338394,
            "uncore": 12.039601711020893
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 726.9945112251366,
            "sha256": "82a7d15255d9e580be3390c3a34d6d3e2feb60a349b88eded778069599535d54",
            "sum": -57824.15603592619,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2100,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 23,
          "gross_j": {
            "core": 3.524589,
            "dram": 0.655699,
            "package-0": 6.557966,
            "psys": 10.241063,
            "uncore": 1.650142
          },
          "net_j": {
            "core": 2.657310072674744,
            "dram": 0.5421522489004915,
            "package-0": 5.241162256291754,
            "psys": 7.905694134036684,
            "uncore": 1.6496929092364736
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.229940307,
          "token_latencies_s": [
            0.016858227,
            0.016458946,
            0.016226336,
            0.016165594,
            0.016198213,
            0.016528615,
            0.016466194,
            0.016565852,
            0.016362389,
            0.016476792,
            0.016638163,
            0.016320655,
            0.016435025,
            0.016206259
          ],
          "watts": {
            "core": 15.328278221355946,
            "dram": 2.851605308155042,
            "package-0": 28.520297661427406,
            "psys": 44.53792000895258,
            "uncore": 7.176392958368973
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.254679,
            "dram": 1.460263,
            "package-0": 18.109451,
            "psys": 29.291917,
            "uncore": 6.407698
          },
          "net_j": {
            "core": 5.413217183586314,
            "dram": 1.0882503205012364,
            "package-0": 13.795213169966411,
            "psys": 21.640559114087203,
            "uncore": 6.406226646071685
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 1900,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 4,
          "gross_j": {
            "core": 1.186887,
            "dram": 0.062805,
            "package-0": 1.419674,
            "psys": 2.024836,
            "uncore": 0.083374
          },
          "logits": {
            "l2": 752.85195897198,
            "sha256": "4970ddf30ee86b82fca779e34e85f3a0321c66b97a1a267f9f290f6c9784648a",
            "sum": -146160.83817428164,
            "top1": 522
          },
          "net_j": {
            "core": 1.0605367438820092,
            "dram": 0.046262846569575175,
            "package-0": 1.227834344921803,
            "psys": 1.6846058329250604,
            "uncore": 0.08330857383639335
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.033499046,
          "watts": {
            "core": 35.430471661790016,
            "dram": 1.8748295100702272,
            "package-0": 42.37953522616734,
            "psys": 60.44458698913397,
            "uncore": 2.4888469958219113
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2718.5155112787256,
          "decode_tokens_per_s": 50.75007511434034,
          "end_to_end_ms": 753.3523319999999,
          "handoff_estimate_ms": 17.1116635,
          "handoff_plus_first_decode_ms": 33.499046,
          "output_tokens_per_s_e2e": 21.238402431865044,
          "prefill_tokens_per_s": 4438.752637374331,
          "steady_tpot_ms": 16.387382499999998,
          "steady_tpot_p95_ms": 28.636520799999996,
          "ttft_ms": 457.786267
        },
        "prefill": {
          "gpu_active_fraction": 0.6086956521739131,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 46,
          "gross_j": {
            "core": 3.780874,
            "dram": 0.726195,
            "package-0": 10.315464,
            "psys": 17.005572,
            "uncore": 4.68706
          },
          "logits": {
            "l2": 658.700794245309,
            "sha256": "d8b9f589d3ed262681597fb65c434b4174e3993033527e14721527676e3f0747",
            "sum": -95245.78444925009,
            "top1": 220
          },
          "net_j": {
            "core": 2.054215512986701,
            "dram": 0.5001357066381704,
            "package-0": 7.693850251333384,
            "psys": 12.356105322470647,
            "uncore": 4.686165908890611
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.457786267,
          "watts": {
            "core": 8.2590376176575,
            "dram": 1.5863188836112465,
            "package-0": 22.53336271443896,
            "psys": 37.147405297765296,
            "uncore": 10.238533433332545
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 726.9945112251366,
            "sha256": "82a7d15255d9e580be3390c3a34d6d3e2feb60a349b88eded778069599535d54",
            "sum": -57824.15603592619,
            "top1": 42
          },
          "gpu_active_fraction": 0.9615384615384616,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 3.286918,
            "dram": 0.671263,
            "package-0": 6.374313,
            "psys": 10.261509,
            "uncore": 1.637264
          },
          "net_j": {
            "core": 2.298464926717604,
            "dram": 0.5418517672934908,
            "package-0": 4.873528573711224,
            "psys": 7.599847958691497,
            "uncore": 1.6367521633446815
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.262067019,
          "token_latencies_s": [
            0.029648017,
            0.028091869,
            0.024304728,
            0.016569629,
            0.016349441,
            0.01634223,
            0.016422857,
            0.01629344,
            0.01620432,
            0.016351908,
            0.016222598,
            0.016349102,
            0.016442159,
            0.016441698
          ],
          "watts": {
            "core": 12.542280263049811,
            "dram": 2.561417314400787,
            "package-0": 24.32321710806349,
            "psys": 39.15604885786868,
            "uncore": 6.247501140156824
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.148967,
            "dram": 1.430722,
            "package-0": 17.74196,
            "psys": 28.552111999999997,
            "uncore": 6.377242
          },
          "net_j": {
            "core": 5.470221093595741,
            "dram": 1.0800125741047237,
            "package-0": 13.674776390598883,
            "psys": 21.338908043878686,
            "uncore": 6.375854902946795
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1750,
          "gpu_sample_count": 3,
          "gross_j": {
            "core": 0.940122,
            "dram": 0.057373,
            "package-0": 1.134763,
            "psys": 1.555477,
            "uncore": 0.067444
          },
          "logits": {
            "l2": 752.85195897198,
            "sha256": "4970ddf30ee86b82fca779e34e85f3a0321c66b97a1a267f9f290f6c9784648a",
            "sum": -146160.83817428164,
            "top1": 522
          },
          "net_j": {
            "core": 0.8369853042785467,
            "dram": 0.043870035325847725,
            "package-0": 0.9781688723661635,
            "psys": 1.2777552422846954,
            "uncore": 0.06739059418559612
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.027344471,
          "watts": {
            "core": 34.38069802118315,
            "dram": 2.098157247218277,
            "package-0": 41.498809759384265,
            "psys": 56.88451606908029,
            "uncore": 2.4664583929965223
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2883.6471589780886,
          "decode_tokens_per_s": 56.11062744373706,
          "end_to_end_ms": 710.2117169999999,
          "handoff_estimate_ms": 10.788283499999999,
          "handoff_plus_first_decode_ms": 27.344471,
          "output_tokens_per_s_e2e": 22.528493429516317,
          "prefill_tokens_per_s": 4588.122524908854,
          "steady_tpot_ms": 16.5561875,
          "steady_tpot_p95_ms": 20.14588655,
          "ttft_ms": 442.88268
        },
        "prefill": {
          "gpu_active_fraction": 0.6363636363636364,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 44,
          "gross_j": {
            "core": 3.656485,
            "dram": 0.713377,
            "package-0": 10.073888,
            "psys": 16.679584,
            "uncore": 4.68883
          },
          "logits": {
            "l2": 658.700794245309,
            "sha256": "d8b9f589d3ed262681597fb65c434b4174e3993033527e14721527676e3f0747",
            "sum": -95245.78444925009,
            "top1": 220
          },
          "net_j": {
            "core": 1.9860392192985992,
            "dram": 0.4946772422744296,
            "package-0": 7.537622920352433,
            "psys": 12.18148431647477,
            "uncore": 4.687965016720171
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.44288268,
          "watts": {
            "core": 8.256102948076453,
            "dram": 1.6107584067184566,
            "package-0": 22.74617738494538,
            "psys": 37.661405047494746,
            "uncore": 10.587070146884047
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 726.9945112251366,
            "sha256": "82a7d15255d9e580be3390c3a34d6d3e2feb60a349b88eded778069599535d54",
            "sum": -57824.15603592619,
            "top1": 42
          },
          "gpu_active_fraction": 0.9583333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 24,
          "gross_j": {
            "core": 3.55236,
            "dram": 0.659972,
            "package-0": 6.533309,
            "psys": 10.317051,
            "uncore": 1.620968
          },
          "net_j": {
            "core": 2.6471965700185955,
            "dram": 0.5414652965044463,
            "package-0": 5.158984597880285,
            "psys": 7.87966848511922,
            "uncore": 1.6204992920410275
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.239984566,
          "token_latencies_s": [
            0.02277963,
            0.016815926,
            0.018727717,
            0.01678692,
            0.016704349,
            0.016471615,
            0.01664076,
            0.016647662,
            0.016404419,
            0.016269311,
            0.016459391,
            0.016440634,
            0.016351194,
            0.016463857
          ],
          "watts": {
            "core": 14.802451921012286,
            "dram": 2.7500601851204047,
            "package-0": 27.223871555139922,
            "psys": 42.990477145934456,
            "uncore": 6.7544677018937955
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 0.9078262252733111,
    "compile_target": "AUTO:GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396367,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03983466699719429,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 12.173551999999999,
            "dram": 2.246699,
            "package-0": 31.7335,
            "psys": 51.562001,
            "uncore": 14.243920999999999
          },
          "net_j": {
            "core": 7.083099794976163,
            "dram": 1.5802417723461608,
            "package-0": 24.004583374046508,
            "psys": 37.854665079881904,
            "uncore": 14.241285083231562
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.791621,
            "dram": 0.078308,
            "package-0": 2.086481,
            "psys": 2.804558,
            "uncore": 0.107849
          },
          "logits": {
            "l2": 731.5050659415324,
            "sha256": "c9bf661180d16881f80683aef6a956fd8906c6d1b49f9d814a4006bfb3fa18cd",
            "sum": -52036.174412652384,
            "top1": 78
          },
          "net_j": {
            "core": 1.5983440114728615,
            "dram": 0.05300359859221261,
            "package-0": 1.79302539529609,
            "psys": 2.2841106014848633,
            "uncore": 0.10774891811444397
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.051243226,
          "watts": {
            "core": 34.96307980297727,
            "dram": 1.5281629614809964,
            "package-0": 40.71720621180251,
            "psys": 54.730316939842936,
            "uncore": 2.1046489149609746
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3034.9201648542553,
          "decode_tokens_per_s": 43.86502438393833,
          "end_to_end_ms": 1349.6236399999998,
          "handoff_estimate_ms": 31.199350500000005,
          "handoff_plus_first_decode_ms": 51.243226,
          "output_tokens_per_s_e2e": 11.855156893961935,
          "prefill_tokens_per_s": 4048.9624797126016,
          "steady_tpot_ms": 20.0438755,
          "steady_tpot_p95_ms": 23.857719749999998,
          "ttft_ms": 1007.6655489999999
        },
        "prefill": {
          "gpu_active_fraction": 0.73,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 100,
          "gross_j": {
            "core": 6.547285,
            "dram": 1.46069,
            "package-0": 22.34705,
            "psys": 37.008938,
            "uncore": 12.173675
          },
          "logits": {
            "l2": 651.1399731232168,
            "sha256": "7cccf7c80f58f7fe8b63aa7d53f62f7cd36a27b775f8f747cddefd8226dc491f",
            "sum": -4677.085442299489,
            "top1": 62
          },
          "net_j": {
            "core": 2.7466155660875904,
            "dram": 0.9630949903358066,
            "package-0": 16.576431596289574,
            "psys": 26.774669893023734,
            "uncore": 12.171706953176527
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.007665549,
          "watts": {
            "core": 6.497478262006157,
            "dram": 1.4495781873753433,
            "package-0": 22.177050730946444,
            "psys": 36.72740229804165,
            "uncore": 12.081066989023359
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 713.6949237157127,
            "sha256": "c6991e9de21ae0c23772f8b7e984041fd8586adc9216fbb0440d1c7c586214f4",
            "sum": 28851.84398850752,
            "top1": 42
          },
          "gpu_active_fraction": 0.9655172413793104,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 3.834646,
            "dram": 0.707701,
            "package-0": 7.299969,
            "psys": 11.748505,
            "uncore": 1.962397
          },
          "net_j": {
            "core": 2.738140217415711,
            "dram": 0.5641431834181415,
            "package-0": 5.635126382460843,
            "psys": 8.79588458537331,
            "uncore": 1.9618292119405916
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.290714865,
          "token_latencies_s": [
            0.02685148,
            0.020833562,
            0.020735698,
            0.022245695,
            0.020263753,
            0.019823998,
            0.019666624,
            0.021071469,
            0.019740248,
            0.021032095,
            0.019667173,
            0.019524074,
            0.019672873,
            0.01955748
          ],
          "watts": {
            "core": 13.19040221765062,
            "dram": 2.434347483400961,
            "package-0": 25.110408440930602,
            "psys": 40.41246738449374,
            "uncore": 6.750246500123067
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.123991,
            "dram": 2.449518,
            "package-0": 33.405189,
            "psys": 55.373577000000004,
            "uncore": 14.503931
          },
          "net_j": {
            "core": 7.641205094654028,
            "dram": 1.731695270873853,
            "package-0": 25.080585716513614,
            "psys": 40.60978290946809,
            "uncore": 14.50109192662821
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 4,
          "gross_j": {
            "core": 1.342343,
            "dram": 0.068054,
            "package-0": 1.592892,
            "psys": 2.169367,
            "uncore": 0.09198
          },
          "logits": {
            "l2": 731.5050659415324,
            "sha256": "c9bf661180d16881f80683aef6a956fd8906c6d1b49f9d814a4006bfb3fa18cd",
            "sum": -52036.174412652384,
            "top1": 78
          },
          "net_j": {
            "core": 1.1885890935430734,
            "dram": 0.04792407870595432,
            "package-0": 1.359444935562888,
            "psys": 1.755345541797223,
            "uncore": 0.09190038379226065
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.040764533,
          "watts": {
            "core": 32.929188714120684,
            "dram": 1.6694414235041035,
            "package-0": 39.075438445474155,
            "psys": 53.21702078618195,
            "uncore": 2.2563732056000743
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2817.7492814720413,
          "decode_tokens_per_s": 46.14865166161512,
          "end_to_end_ms": 1453.6424610000001,
          "handoff_estimate_ms": 21.256463499999995,
          "handoff_plus_first_decode_ms": 40.764533,
          "output_tokens_per_s_e2e": 11.006833130750161,
          "prefill_tokens_per_s": 3615.079520598528,
          "steady_tpot_ms": 19.5080695,
          "steady_tpot_p95_ms": 23.44630155,
          "ttft_ms": 1128.605879
        },
        "prefill": {
          "gpu_active_fraction": 0.7053571428571429,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 112,
          "gross_j": {
            "core": 8.214579,
            "dram": 1.676143,
            "package-0": 24.804624,
            "psys": 41.804459,
            "uncore": 12.45999
          },
          "logits": {
            "l2": 651.1399731232168,
            "sha256": "7cccf7c80f58f7fe8b63aa7d53f62f7cd36a27b775f8f747cddefd8226dc491f",
            "sum": -4677.085442299489,
            "top1": 62
          },
          "net_j": {
            "core": 3.957752048333375,
            "dram": 1.118826483795405,
            "package-0": 18.341414195560034,
            "psys": 30.341870878119476,
            "uncore": 12.45778574758974
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.128605879,
          "watts": {
            "core": 7.27851870422518,
            "dram": 1.4851446649251416,
            "package-0": 21.978109862388905,
            "psys": 37.040795000147256,
            "uncore": 11.040160459770208
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 713.6949237157127,
            "sha256": "c6991e9de21ae0c23772f8b7e984041fd8586adc9216fbb0440d1c7c586214f4",
            "sum": 28851.84398850752,
            "top1": 42
          },
          "gpu_active_fraction": 0.9655172413793104,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 3.567069,
            "dram": 0.705321,
            "package-0": 7.007673,
            "psys": 11.399751,
            "uncore": 1.951961
          },
          "net_j": {
            "core": 2.49486395277758,
            "dram": 0.5649447083724938,
            "package-0": 5.37972658539069,
            "psys": 8.51256648955139,
            "uncore": 1.951405795246209
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.284272049,
          "token_latencies_s": [
            0.026313824,
            0.020670523,
            0.021902251,
            0.02035076,
            0.019545107,
            0.019262625,
            0.019415048,
            0.019098347,
            0.019386232,
            0.019517558,
            0.019344895,
            0.020663806,
            0.019271766,
            0.019498581
          ],
          "watts": {
            "core": 12.548082066274478,
            "dram": 2.4811479091284134,
            "package-0": 24.65129098921716,
            "psys": 40.10155426853098,
            "uncore": 6.866524538260179
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.166531,
            "dram": 2.441583,
            "package-0": 33.326574,
            "psys": 54.908367999999996,
            "uncore": 14.399315999999999
          },
          "net_j": {
            "core": 7.737410665675533,
            "dram": 1.7307863279396978,
            "package-0": 25.08345202846858,
            "psys": 40.289082094538706,
            "uncore": 14.396504715510778
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2000,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.592586,
            "dram": 0.070861,
            "package-0": 1.877009,
            "psys": 2.525018,
            "uncore": 0.111755
          },
          "logits": {
            "l2": 731.5050659415324,
            "sha256": "c9bf661180d16881f80683aef6a956fd8906c6d1b49f9d814a4006bfb3fa18cd",
            "sum": -52036.174412652384,
            "top1": 78
          },
          "net_j": {
            "core": 1.4165011921403539,
            "dram": 0.04780745030508654,
            "package-0": 1.6096565636680942,
            "psys": 2.0508649160402777,
            "uncore": 0.1116638204978634
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.04668509,
          "watts": {
            "core": 34.11337538387524,
            "dram": 1.517850774197929,
            "package-0": 40.20574877332356,
            "psys": 54.086176121755365,
            "uncore": 2.393804959999006
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2845.602066983094,
          "decode_tokens_per_s": 44.477808863252505,
          "end_to_end_ms": 1439.414192,
          "handoff_estimate_ms": 26.82702,
          "handoff_plus_first_decode_ms": 46.685089999999995,
          "output_tokens_per_s_e2e": 11.115633074152711,
          "prefill_tokens_per_s": 3701.797147118028,
          "steady_tpot_ms": 19.858069999999998,
          "steady_tpot_p95_ms": 23.73208415,
          "ttft_ms": 1102.167363
        },
        "prefill": {
          "gpu_active_fraction": 0.6972477064220184,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 109,
          "gross_j": {
            "core": 7.796001,
            "dram": 1.65881,
            "package-0": 24.176146,
            "psys": 40.703326,
            "uncore": 12.315093
          },
          "logits": {
            "l2": 651.1399731232168,
            "sha256": "7cccf7c80f58f7fe8b63aa7d53f62f7cd36a27b775f8f747cddefd8226dc491f",
            "sum": -4677.085442299489,
            "top1": 62
          },
          "net_j": {
            "core": 3.6388937020805168,
            "dram": 1.1145490791142907,
            "package-0": 17.86434217262038,
            "psys": 29.509258382133034,
            "uncore": 12.312940384005694
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.102167363,
          "watts": {
            "core": 7.073336828610121,
            "dram": 1.505043658237955,
            "package-0": 21.93509516938944,
            "psys": 36.93025883946447,
            "uncore": 11.173523562228725
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 713.6949237157127,
            "sha256": "c6991e9de21ae0c23772f8b7e984041fd8586adc9216fbb0440d1c7c586214f4",
            "sum": 28851.84398850752,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 3.777944,
            "dram": 0.711912,
            "package-0": 7.273419,
            "psys": 11.680024,
            "uncore": 1.972468
          },
          "net_j": {
            "core": 2.6820157714546626,
            "dram": 0.5684297985203205,
            "package-0": 5.609453292180108,
            "psys": 8.728958796365394,
            "uncore": 1.9719005110072196
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.290561739,
          "token_latencies_s": [
            0.028723609,
            0.020800673,
            0.020679015,
            0.020717207,
            0.019930645,
            0.019642804,
            0.019785495,
            0.019738085,
            0.019739745,
            0.019631231,
            0.019536176,
            0.02104434,
            0.019687741,
            0.020872349
          ],
          "watts": {
            "core": 13.002207424150916,
            "dram": 2.4501230012255677,
            "package-0": 25.032266894575546,
            "psys": 40.198079899294655,
            "uncore": 6.788464327025522
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.790025919675827,
    "compile_target": "NPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396436,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 4096,
      "NPUW_LLM_MIN_RESPONSE_LEN": 4096,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03980210516601801,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.947664,
            "dram": 2.564935,
            "package-0": 28.757006,
            "psys": 48.551816,
            "uncore": 10.149876
          },
          "net_j": {
            "core": 7.202749779567352,
            "dram": 1.9437166387542306,
            "package-0": 21.552725332983442,
            "psys": 35.77492889605537,
            "uncore": 10.14836493559327
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 18,
          "gross_j": {
            "core": 2.852715,
            "dram": 0.15979,
            "package-0": 3.322623,
            "psys": 5.06798,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1194.5389349351794,
            "sha256": "e1f0572991cfbd0bb62604abe32d92b58e7c7dae72ac6196e0b2ea2540eacc99",
            "sum": -145173.30570931733,
            "top1": 151643
          },
          "net_j": {
            "core": 2.1830452256811927,
            "dram": 0.07211483023892871,
            "package-0": 2.3058524569973446,
            "psys": 3.2647239861082875,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 27636,
          "npu_utilization": 0.1556532431664968,
          "seconds": 0.177548501,
          "watts": {
            "core": 16.067243507733135,
            "dram": 0.8999794371679882,
            "package-0": 18.713889338891125,
            "psys": 28.544200437941182,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3255.93157800964,
          "decode_tokens_per_s": 30.970805863321115,
          "end_to_end_ms": 1258.0116939999998,
          "handoff_estimate_ms": 157.0572,
          "handoff_plus_first_decode_ms": 177.548501,
          "output_tokens_per_s_e2e": 12.718482726600156,
          "prefill_tokens_per_s": 5273.466644476229,
          "steady_tpot_ms": 20.491301,
          "steady_tpot_p95_ms": 28.679078849999996,
          "ttft_ms": 773.684613
        },
        "prefill": {
          "gpu_active_fraction": 0.987012987012987,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 77,
          "gross_j": {
            "core": 5.708116,
            "dram": 1.624813,
            "package-0": 19.831736,
            "psys": 34.060459,
            "uncore": 10.149876
          },
          "logits": {
            "l2": 1316.4831161793027,
            "sha256": "e44a9a14f0e2b695504bf9bdfb6472c0d1fa984ec295e707921b018c2f670cae",
            "sum": 359625.1051243767,
            "top1": 89372
          },
          "net_j": {
            "core": 2.7899657635097266,
            "dram": 1.2427600431783385,
            "package-0": 15.401060890707981,
            "psys": 26.202598058398856,
            "uncore": 10.14836493559327
          },
          "npu_busy_delta_us": 5404,
          "npu_utilization": 0.006984758271262143,
          "seconds": 0.773684613,
          "watts": {
            "core": 7.377833168823793,
            "dram": 2.10009734289494,
            "package-0": 25.63284272011236,
            "psys": 44.023699615698575,
            "uncore": 13.118880522443582
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1245.6294987734661,
            "sha256": "5a20b8596177229ad661bfa977ae7e167f5368166f38b44599912f24e0db964d",
            "sum": -187297.29206892848,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 30,
          "gross_j": {
            "core": 3.386833,
            "dram": 0.780332,
            "package-0": 5.602647,
            "psys": 9.423377,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.229738790376433,
            "dram": 0.6288417653369635,
            "package-0": 3.8458119852781154,
            "psys": 6.307606851548226,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 287678,
          "npu_utilization": 0.9377382214886057,
          "seconds": 0.30677858,
          "token_latencies_s": [
            0.033055289,
            0.026322658,
            0.023812932,
            0.022443461,
            0.021143438,
            0.02091615,
            0.020546206,
            0.020436396,
            0.020137225,
            0.019855986,
            0.01985136,
            0.019418918,
            0.019410153,
            0.01938896
          ],
          "watts": {
            "core": 11.039991775175437,
            "dram": 2.543632609551814,
            "package-0": 18.262836342745963,
            "psys": 30.717193488541472,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.995272,
            "dram": 2.47503,
            "package-0": 27.288567,
            "psys": 45.607854,
            "uncore": 9.983251000000001
          },
          "net_j": {
            "core": 6.740093321162535,
            "dram": 1.9179292806039978,
            "package-0": 20.82785979508564,
            "psys": 34.14970427149738,
            "uncore": 9.981628450628
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.16666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.945124,
            "dram": 0.086181,
            "package-0": 2.185236,
            "psys": 2.944329,
            "uncore": 0.010804
          },
          "logits": {
            "l2": 1076.3716215015206,
            "sha256": "938aa50fdd336de86a3bb8faeb3e45e76c6d3a3148e88756f8b550ab3d963cd8",
            "sum": -83517.90235760808,
            "top1": 151643
          },
          "net_j": {
            "core": 1.7148955245361954,
            "dram": 0.056038799404707546,
            "package-0": 1.8356763498722581,
            "psys": 2.3243803344483394,
            "uncore": 0.010684784049313386
          },
          "npu_busy_delta_us": 25329,
          "npu_utilization": 0.414956616399615,
          "seconds": 0.061040116,
          "watts": {
            "core": 31.86632214132752,
            "dram": 1.4118747742877815,
            "package-0": 35.799997496728224,
            "psys": 48.235966655109245,
            "uncore": 0.17699835301754668
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3630.6621205092188,
          "decode_tokens_per_s": 41.8477013338037,
          "end_to_end_ms": 1128.168875,
          "handoff_estimate_ms": 40.7916245,
          "handoff_plus_first_decode_ms": 61.040116,
          "output_tokens_per_s_e2e": 14.182273908239136,
          "prefill_tokens_per_s": 5300.585773969865,
          "steady_tpot_ms": 20.2484915,
          "steady_tpot_p95_ms": 26.01549035,
          "ttft_ms": 769.726248
        },
        "prefill": {
          "gpu_active_fraction": 0.9868421052631579,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 76,
          "gross_j": {
            "core": 5.739915,
            "dram": 1.618404,
            "package-0": 19.639599,
            "psys": 33.500768,
            "uncore": 9.972447
          },
          "logits": {
            "l2": 1334.0000819701768,
            "sha256": "4ec570d789e7f405abaa32f13c5b5ba60aa64d12f6cb3a84109463b050062787",
            "sum": 371782.7636419125,
            "top1": 89372
          },
          "net_j": {
            "core": 2.836694753801345,
            "dram": 1.238305722161245,
            "package-0": 15.231592338528067,
            "psys": 25.683109850600566,
            "uncore": 9.970943666578686
          },
          "npu_busy_delta_us": 5632,
          "npu_utilization": 0.007316887029166244,
          "seconds": 0.769726248,
          "watts": {
            "core": 7.457086223724567,
            "dram": 2.102570886994099,
            "package-0": 25.515043888694308,
            "psys": 43.52296428379041,
            "uncore": 12.955835955850112
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1276.7456685137324,
            "sha256": "c285e1961626b95265953e042fd1dcac6f2895b14eba95347b13315d1aba89d0",
            "sum": -156027.74945994467,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 30,
          "gross_j": {
            "core": 3.310233,
            "dram": 0.770445,
            "package-0": 5.463732,
            "psys": 9.162757,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.188503042824994,
            "dram": 0.6235847590380453,
            "package-0": 3.760591106685316,
            "psys": 6.142214086448472,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 287886,
          "npu_utilization": 0.9680012419262997,
          "seconds": 0.297402511,
          "token_latencies_s": [
            0.025895454,
            0.026238415,
            0.024245749,
            0.022298555,
            0.021324219,
            0.020759729,
            0.020372359,
            0.020124624,
            0.019862358,
            0.019490558,
            0.019343102,
            0.019181057,
            0.019005239,
            0.019225364
          ],
          "watts": {
            "core": 11.130481006597824,
            "dram": 2.590580010267634,
            "package-0": 18.371505948717427,
            "psys": 30.809279212844306,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.950778,
            "dram": 2.4689870000000003,
            "package-0": 27.393546,
            "psys": 45.691716,
            "uncore": 10.064854
          },
          "net_j": {
            "core": 6.695186313787788,
            "dram": 1.9118322084432375,
            "package-0": 20.932211719250972,
            "psys": 34.23245414420591,
            "uncore": 10.062650387229755
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.3333333333333333,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.853084,
            "dram": 0.083801,
            "package-0": 2.117304,
            "psys": 2.851555,
            "uncore": 0.038147
          },
          "logits": {
            "l2": 1132.7733093910065,
            "sha256": "ba6f530cba6ce38dd32d9851f295a6bd9ea606a03641d5c0ab35981cde51d82e",
            "sum": -121606.2538428083,
            "top1": 51186
          },
          "net_j": {
            "core": 1.6358691832395387,
            "dram": 0.05536258652454415,
            "package-0": 1.7875032001797981,
            "psys": 2.266648916956126,
            "uncore": 0.038034522727885205
          },
          "npu_busy_delta_us": 26224,
          "npu_utilization": 0.4553582164904999,
          "seconds": 0.057589825,
          "watts": {
            "core": 32.17728131662147,
            "dram": 1.4551355209014787,
            "package-0": 36.76524455491921,
            "psys": 49.514909968905094,
            "uncore": 0.6623913165216252
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3630.309762872128,
          "decode_tokens_per_s": 41.678565086582346,
          "end_to_end_ms": 1128.278375,
          "handoff_estimate_ms": 37.2021285,
          "handoff_plus_first_decode_ms": 57.589825,
          "output_tokens_per_s_e2e": 14.18089751121925,
          "prefill_tokens_per_s": 5309.864786011121,
          "steady_tpot_ms": 20.3876965,
          "steady_tpot_p95_ms": 26.38795515,
          "ttft_ms": 768.381148
        },
        "prefill": {
          "gpu_active_fraction": 0.9868421052631579,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 76,
          "gross_j": {
            "core": 5.631028,
            "dram": 1.610591,
            "package-0": 19.576,
            "psys": 33.383825,
            "uncore": 9.989476
          },
          "logits": {
            "l2": 1341.2597838450145,
            "sha256": "55370d341ddc06cd6b806e3e24b2ff4f1e4aa07f6b532c5ea1c0e77cc32101c3",
            "sum": 376821.0504594818,
            "top1": 89372
          },
          "net_j": {
            "core": 2.732881143936225,
            "dram": 1.2311569455716451,
            "package-0": 15.17569634953458,
            "psys": 25.579828242561774,
            "uncore": 9.987975293660442
          },
          "npu_busy_delta_us": 4538,
          "npu_utilization": 0.005905923136989822,
          "seconds": 0.768381148,
          "watts": {
            "core": 7.3284307074124095,
            "dram": 2.0960834400898136,
            "package-0": 25.47693947327297,
            "psys": 43.44695999751415,
            "uncore": 13.00067814781942
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1295.5273517640333,
            "sha256": "d5f59dc8ca74197855cb3401a475dc3f2f32742d13458fbb24ee3aea7556bd73",
            "sum": -187058.67358023673,
            "top1": 42
          },
          "gpu_active_fraction": 0.06666666666666667,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 30,
          "gross_j": {
            "core": 3.466666,
            "dram": 0.774595,
            "package-0": 5.700242,
            "psys": 9.456336,
            "uncore": 0.037231
          },
          "net_j": {
            "core": 2.326435986612025,
            "dram": 0.6253126763470483,
            "package-0": 3.969012169536594,
            "psys": 6.385976984688012,
            "uncore": 0.036640570841427786
          },
          "npu_busy_delta_us": 290036,
          "npu_utilization": 0.9594075371002659,
          "seconds": 0.302307402,
          "token_latencies_s": [
            0.026277177,
            0.026593686,
            0.024474226,
            0.022381668,
            0.021525552,
            0.020739282,
            0.023106132,
            0.020036111,
            0.020015511,
            0.020003811,
            0.01951297,
            0.019395297,
            0.019172401,
            0.019045849
          ],
          "watts": {
            "core": 11.467354014705865,
            "dram": 2.5622759974630065,
            "package-0": 18.85578044827364,
            "psys": 31.280530802219655,
            "uncore": 0.12315609791122481
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.350780539214611,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396280,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.040753516368567944,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.879669,
            "dram": 2.247492,
            "package-0": 31.45158,
            "psys": 51.033440000000006,
            "uncore": 14.252771
          },
          "net_j": {
            "core": 6.801667613264014,
            "dram": 1.5826648707154083,
            "package-0": 23.741567654569312,
            "psys": 37.359631071567144,
            "uncore": 14.250141530462455
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.674007,
            "dram": 0.075989,
            "package-0": 1.973384,
            "psys": 2.661859,
            "uncore": 0.117066
          },
          "logits": {
            "l2": 731.5050659415324,
            "sha256": "c9bf661180d16881f80683aef6a956fd8906c6d1b49f9d814a4006bfb3fa18cd",
            "sum": -52036.174412652384,
            "top1": 78
          },
          "net_j": {
            "core": 1.483971391415538,
            "dram": 0.05110896974891198,
            "package-0": 1.6848498353412904,
            "psys": 2.15013984053505,
            "uncore": 0.11696759655231151
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.050383844,
          "watts": {
            "core": 33.22507508557704,
            "dram": 1.508201716407347,
            "package-0": 39.1669996437747,
            "psys": 52.831598160712,
            "uncore": 2.323482900590118
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3042.3615254630645,
          "decode_tokens_per_s": 44.8805488036593,
          "end_to_end_ms": 1346.3225740000003,
          "handoff_estimate_ms": 30.657608499999995,
          "handoff_plus_first_decode_ms": 50.383843999999996,
          "output_tokens_per_s_e2e": 11.884224708840096,
          "prefill_tokens_per_s": 4031.2139982578155,
          "steady_tpot_ms": 19.7262355,
          "steady_tpot_p95_ms": 22.792098749999997,
          "ttft_ms": 1012.102062
        },
        "prefill": {
          "gpu_active_fraction": 0.74,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 100,
          "gross_j": {
            "core": 6.873701,
            "dram": 1.468319,
            "package-0": 22.724368,
            "psys": 37.347683,
            "uncore": 12.197234
          },
          "logits": {
            "l2": 651.1399731232168,
            "sha256": "7cccf7c80f58f7fe8b63aa7d53f62f7cd36a27b775f8f747cddefd8226dc491f",
            "sum": -4677.085442299489,
            "top1": 62
          },
          "net_j": {
            "core": 3.0562981178030983,
            "dram": 0.968533197238532,
            "package-0": 16.928342928778207,
            "psys": 27.06835583196211,
            "uncore": 12.195257288332002
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.012102062,
          "watts": {
            "core": 6.791509728195771,
            "dram": 1.4507617908597836,
            "package-0": 22.452644701755382,
            "psys": 36.90110355688614,
            "uncore": 12.05138736294759
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 713.6949237157127,
            "sha256": "c6991e9de21ae0c23772f8b7e984041fd8586adc9216fbb0440d1c7c586214f4",
            "sum": 28851.84398850752,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 28,
          "gross_j": {
            "core": 3.331961,
            "dram": 0.703184,
            "package-0": 6.753828,
            "psys": 11.023898,
            "uncore": 1.938471
          },
          "net_j": {
            "core": 2.2613981040453774,
            "dram": 0.5630227037279644,
            "package-0": 5.128374890449814,
            "psys": 8.141135399069984,
            "uncore": 1.9379166455781418
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.283836668,
          "token_latencies_s": [
            0.02613058,
            0.020524535,
            0.020036894,
            0.020026238,
            0.019896589,
            0.020994455,
            0.019629611,
            0.019503504,
            0.019547554,
            0.019274896,
            0.01982286,
            0.019396263,
            0.019531048,
            0.01949757
          ],
          "watts": {
            "core": 11.73900829472815,
            "dram": 2.4774247984055395,
            "package-0": 23.79476918042175,
            "psys": 38.83887898514931,
            "uncore": 6.829529861871124
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.175198,
            "dram": 2.480158,
            "package-0": 33.601415,
            "psys": 55.254863,
            "uncore": 14.557091
          },
          "net_j": {
            "core": 7.552141487204246,
            "dram": 1.7439706234128591,
            "package-0": 25.063836564918784,
            "psys": 40.113354659615084,
            "uncore": 14.554179292286229
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2000,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.860774,
            "dram": 0.075928,
            "package-0": 2.137995,
            "psys": 2.824761,
            "uncore": 0.089965
          },
          "logits": {
            "l2": 731.5050659415324,
            "sha256": "c9bf661180d16881f80683aef6a956fd8906c6d1b49f9d814a4006bfb3fa18cd",
            "sum": -52036.174412652384,
            "top1": 78
          },
          "net_j": {
            "core": 1.664254997875035,
            "dram": 0.05019914447548576,
            "package-0": 1.8396169931619377,
            "psys": 2.2955836561571883,
            "uncore": 0.08986323934845977
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.052102776,
          "watts": {
            "core": 35.71352896820699,
            "dram": 1.4572736009305913,
            "package-0": 41.034185971204295,
            "psys": 54.21517271939599,
            "uncore": 1.726683430456757
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2747.4587904457026,
          "decode_tokens_per_s": 42.67533289256163,
          "end_to_end_ms": 1490.8321879999999,
          "handoff_estimate_ms": 32.602312500000004,
          "handoff_plus_first_decode_ms": 52.102776000000006,
          "output_tokens_per_s_e2e": 10.732260900178526,
          "prefill_tokens_per_s": 3581.0171932174894,
          "steady_tpot_ms": 19.500463500000002,
          "steady_tpot_p95_ms": 32.233098700000006,
          "ttft_ms": 1139.341081
        },
        "prefill": {
          "gpu_active_fraction": 0.6902654867256637,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 113,
          "gross_j": {
            "core": 7.796245,
            "dram": 1.692867,
            "package-0": 24.510069,
            "psys": 41.102189,
            "uncore": 12.54904
          },
          "logits": {
            "l2": 651.1399731232168,
            "sha256": "7cccf7c80f58f7fe8b63aa7d53f62f7cd36a27b775f8f747cddefd8226dc491f",
            "sum": -4677.085442299489,
            "top1": 62
          },
          "net_j": {
            "core": 3.4989274766860987,
            "dram": 1.130249337052581,
            "package-0": 17.98538170008525,
            "psys": 29.530569726305686,
            "uncore": 12.546814780930498
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.139341081,
          "watts": {
            "core": 6.842766516552913,
            "dram": 1.4858298609878704,
            "package-0": 21.512494729398774,
            "psys": 36.07540330585166,
            "uncore": 11.014296078032844
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 713.6949237157127,
            "sha256": "c6991e9de21ae0c23772f8b7e984041fd8586adc9216fbb0440d1c7c586214f4",
            "sum": 28851.84398850752,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 30,
          "gross_j": {
            "core": 3.518179,
            "dram": 0.711363,
            "package-0": 6.953351,
            "psys": 11.327913,
            "uncore": 1.918086
          },
          "net_j": {
            "core": 2.388959012643112,
            "dram": 0.5635221418847923,
            "package-0": 5.238837871671596,
            "psys": 8.28720127715221,
            "uncore": 1.9175012720072706
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.299388331,
          "token_latencies_s": [
            0.033592562,
            0.03150108,
            0.019775257,
            0.019514018,
            0.019912938,
            0.019486909,
            0.019429031,
            0.019411923,
            0.019301678,
            0.019409982,
            0.019374999,
            0.019550814,
            0.019701982,
            0.019399599
          ],
          "watts": {
            "core": 11.751222862456853,
            "dram": 2.3760545296603426,
            "package-0": 23.225190430017125,
            "psys": 37.83685543843057,
            "uncore": 6.406682563723567
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.194912,
            "dram": 2.443354,
            "package-0": 33.474402000000005,
            "psys": 54.61656,
            "uncore": 14.518579
          },
          "net_j": {
            "core": 7.731456499365104,
            "dram": 1.7280620652737015,
            "package-0": 25.1791483853329,
            "psys": 39.90481793553073,
            "uncore": 14.515749936217901
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.600154,
            "dram": 0.075684,
            "package-0": 1.877559,
            "psys": 2.553644,
            "uncore": 0.096192
          },
          "logits": {
            "l2": 731.5050659415324,
            "sha256": "c9bf661180d16881f80683aef6a956fd8906c6d1b49f9d814a4006bfb3fa18cd",
            "sum": -52036.174412652384,
            "top1": 78
          },
          "net_j": {
            "core": 1.3913281928071148,
            "dram": 0.04834390074771877,
            "package-0": 1.5604953702603133,
            "psys": 1.991327456656816,
            "uncore": 0.0960838666898947
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.05536566,
          "watts": {
            "core": 28.901561003697964,
            "dram": 1.3669845171176502,
            "package-0": 33.911977207532615,
            "psys": 46.12324679232578,
            "uncore": 1.7373946233098279
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2827.7188390128445,
          "decode_tokens_per_s": 42.97830372995891,
          "end_to_end_ms": 1448.517421,
          "handoff_estimate_ms": 35.464899499999994,
          "handoff_plus_first_decode_ms": 55.36566,
          "output_tokens_per_s_e2e": 11.045776714893924,
          "prefill_tokens_per_s": 3710.76374837605,
          "steady_tpot_ms": 19.9007605,
          "steady_tpot_p95_ms": 24.891375349999993,
          "ttft_ms": 1099.504112
        },
        "prefill": {
          "gpu_active_fraction": 0.6972477064220184,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 109,
          "gross_j": {
            "core": 8.526407,
            "dram": 1.658993,
            "package-0": 25.030515,
            "psys": 41.232133,
            "uncore": 12.435454
          },
          "logits": {
            "l2": 651.1399731232168,
            "sha256": "7cccf7c80f58f7fe8b63aa7d53f62f7cd36a27b775f8f747cddefd8226dc491f",
            "sum": -4677.085442299489,
            "top1": 62
          },
          "net_j": {
            "core": 4.379344837275009,
            "dram": 1.1160472182578274,
            "package-0": 18.73396286525744,
            "psys": 30.065114460864784,
            "uncore": 12.433306585535782
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.099504112,
          "watts": {
            "core": 7.7547749998774,
            "dram": 1.5088556576494185,
            "package-0": 22.76527638852523,
            "psys": 37.50066284426956,
            "uncore": 11.310056837695575
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 713.6949237157127,
            "sha256": "c6991e9de21ae0c23772f8b7e984041fd8586adc9216fbb0440d1c7c586214f4",
            "sum": 28851.84398850752,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 3.068351,
            "dram": 0.708677,
            "package-0": 6.566328,
            "psys": 10.830783,
            "uncore": 1.986933
          },
          "net_j": {
            "core": 1.9607834692829802,
            "dram": 0.5636709462681554,
            "package-0": 4.884690149815148,
            "psys": 7.848376018009131,
            "uncore": 1.9863594839922254
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.293647649,
          "token_latencies_s": [
            0.031669173,
            0.020945545,
            0.020480787,
            0.020005821,
            0.019894239,
            0.019690039,
            0.019767622,
            0.019636265,
            0.019509863,
            0.019850594,
            0.021241792,
            0.019907282,
            0.021150328,
            0.019873929
          ],
          "watts": {
            "core": 10.449090978419513,
            "dram": 2.4133583306842685,
            "package-0": 22.361248327242695,
            "psys": 36.883601952488306,
            "uncore": 6.766384838313485
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 0.9227810930460691,
    "compile_target": "AUTO:GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396852,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.039894359186291695,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 22.641299,
            "dram": 4.593249999999999,
            "package-0": 71.594605,
            "psys": 115.939156,
            "uncore": 37.757960000000004
          },
          "net_j": {
            "core": 11.152216496772057,
            "dram": 3.0890649561190755,
            "package-0": 54.15054360826083,
            "psys": 85.00188226815493,
            "uncore": 37.75201077112907
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.093134,
            "dram": 0.095886,
            "package-0": 2.497369,
            "psys": 3.405509,
            "uncore": 0.168029
          },
          "logits": {
            "l2": 751.8547055410498,
            "sha256": "87ebcfb2f1baeab184e419908d200db26fa2774137672858533c8384b10828b9",
            "sum": 20114.015098012285,
            "top1": 78
          },
          "net_j": {
            "core": 1.8299457595340742,
            "dram": 0.06142860839166613,
            "package-0": 2.097766000980581,
            "psys": 2.696807792918463,
            "uncore": 0.1678927169610168
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.069778687,
          "watts": {
            "core": 29.996752446775044,
            "dram": 1.3741445149290354,
            "package-0": 35.789853712781955,
            "psys": 48.80442935247549,
            "uncore": 2.408027539985096
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2689.3559239239694,
          "decode_tokens_per_s": 37.56383805239734,
          "end_to_end_ms": 3046.082494,
          "handoff_estimate_ms": 46.99103600000001,
          "handoff_plus_first_decode_ms": 69.778687,
          "output_tokens_per_s_e2e": 5.252648288914003,
          "prefill_tokens_per_s": 3089.0571753033405,
          "steady_tpot_ms": 22.787651,
          "steady_tpot_p95_ms": 26.263037849999996,
          "ttft_ms": 2646.762276
        },
        "prefill": {
          "gpu_active_fraction": 0.7977099236641222,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 262,
          "gross_j": {
            "core": 17.217058,
            "dram": 3.733999,
            "package-0": 61.81869,
            "psys": 100.485582,
            "uncore": 35.272798
          },
          "logits": {
            "l2": 655.8555450667305,
            "sha256": "e30c9d66db32ee71e32369110e91308f4656be42fddef82b606d730823ce7e00",
            "sum": 5070.977332231589,
            "top1": 8
          },
          "net_j": {
            "core": 7.234114263153401,
            "dram": 2.4270021481059962,
            "package-0": 46.66142366764463,
            "psys": 73.60396956927202,
            "uncore": 35.26762867363478
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.646762276,
          "watts": {
            "core": 6.504950654661666,
            "dram": 1.4107798927991066,
            "package-0": 23.356343922743743,
            "psys": 37.965473103183975,
            "uncore": 13.326772230299085
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 724.3605232286193,
            "sha256": "546555705e83653d0439120b2a2a1aa452842d59c8f62055b758ef9c1470d280",
            "sum": 84226.24423345178,
            "top1": 42
          },
          "gpu_active_fraction": 0.9696969696969697,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 3.331107,
            "dram": 0.763365,
            "package-0": 7.278546,
            "psys": 12.048065,
            "uncore": 2.317133
          },
          "net_j": {
            "core": 2.0881564740845815,
            "dram": 0.6006341996214131,
            "package-0": 5.391353939635625,
            "psys": 8.701104905964442,
            "uncore": 2.3164893805332727
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.329541531,
          "token_latencies_s": [
            0.029823589,
            0.024345818,
            0.024109697,
            0.023475631,
            0.023261902,
            0.022716271,
            0.022857281,
            0.022666854,
            0.022666097,
            0.022701305,
            0.022651963,
            0.022655375,
            0.022814111,
            0.022761191
          ],
          "watts": {
            "core": 10.108307107427986,
            "dram": 2.3164455104749755,
            "package-0": 22.086885309760852,
            "psys": 36.560080799041984,
            "uncore": 7.031383853102267
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 24.198791,
            "dram": 4.986986,
            "package-0": 74.670951,
            "psys": 122.276848,
            "uncore": 38.38759700000001
          },
          "net_j": {
            "core": 11.920416480937828,
            "dram": 3.3794644847220856,
            "package-0": 56.02849466450851,
            "psys": 89.21420497592791,
            "uncore": 38.38123906322338
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.63702,
            "dram": 0.081848,
            "package-0": 1.943293,
            "psys": 2.64721,
            "uncore": 0.115356
          },
          "logits": {
            "l2": 751.8547055410498,
            "sha256": "87ebcfb2f1baeab184e419908d200db26fa2774137672858533c8384b10828b9",
            "sum": 20114.015098012285,
            "top1": 78
          },
          "net_j": {
            "core": 1.4296665622122018,
            "dram": 0.0547006677539247,
            "package-0": 1.628464892747548,
            "psys": 2.088858185372689,
            "uncore": 0.11524862910609032
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.054975293,
          "watts": {
            "core": 29.77737653894814,
            "dram": 1.4888142569790397,
            "package-0": 35.34847917954707,
            "psys": 48.152721987311644,
            "uncore": 2.0983244236642813
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2516.4757796349845,
          "decode_tokens_per_s": 39.09940484893042,
          "end_to_end_ms": 3255.3462529999997,
          "handoff_estimate_ms": 32.187742500000006,
          "handoff_plus_first_decode_ms": 54.975293,
          "output_tokens_per_s_e2e": 4.914991757099579,
          "prefill_tokens_per_s": 2847.0854303571946,
          "steady_tpot_ms": 22.7875505,
          "steady_tpot_p95_ms": 26.0898805,
          "ttft_ms": 2871.7087
        },
        "prefill": {
          "gpu_active_fraction": 0.7640845070422535,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 284,
          "gross_j": {
            "core": 19.335217,
            "dram": 4.145497,
            "package-0": 65.556351,
            "psys": 107.772978,
            "uncore": 35.956878
          },
          "logits": {
            "l2": 655.8555450667305,
            "sha256": "e30c9d66db32ee71e32369110e91308f4656be42fddef82b606d730823ce7e00",
            "sum": 5070.977332231589,
            "top1": 8
          },
          "net_j": {
            "core": 8.503830048226439,
            "dram": 2.7274194248091774,
            "package-0": 49.11087949334702,
            "psys": 78.6067166446301,
            "uncore": 35.951269336301436
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.8717087,
          "watts": {
            "core": 6.733000808891235,
            "dram": 1.4435645927457752,
            "package-0": 22.82834293046506,
            "psys": 37.52921666462897,
            "uncore": 12.521074299771422
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 724.3605232286193,
            "sha256": "546555705e83653d0439120b2a2a1aa452842d59c8f62055b758ef9c1470d280",
            "sum": 84226.24423345178,
            "top1": 42
          },
          "gpu_active_fraction": 0.9696969696969697,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 3.226554,
            "dram": 0.759641,
            "package-0": 7.171307,
            "psys": 11.85666,
            "uncore": 2.315363
          },
          "net_j": {
            "core": 1.986919870499188,
            "dram": 0.5973443921589834,
            "package-0": 5.289150278413937,
            "psys": 8.518630145925131,
            "uncore": 2.3147210978158577
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.32866226,
          "token_latencies_s": [
            0.029580556,
            0.02420987,
            0.024210286,
            0.023369866,
            0.022887726,
            0.022622746,
            0.022806878,
            0.022566795,
            0.022561561,
            0.022768223,
            0.023077653,
            0.022547966,
            0.022693308,
            0.022734013
          ],
          "watts": {
            "core": 9.817233046471475,
            "dram": 2.3113119224580276,
            "package-0": 21.81968504689282,
            "psys": 36.07551411591949,
            "uncore": 7.044809464889581
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 26.123528999999998,
            "dram": 4.96721,
            "package-0": 76.570728,
            "psys": 124.765123,
            "uncore": 38.406639999999996
          },
          "net_j": {
            "core": 13.912156539102966,
            "dram": 3.368460594572713,
            "package-0": 58.030001982458366,
            "psys": 91.88290004110453,
            "uncore": 38.4003167579502
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.868587,
            "dram": 0.082092,
            "package-0": 2.178766,
            "psys": 2.918755,
            "uncore": 0.112304
          },
          "logits": {
            "l2": 751.8547055410498,
            "sha256": "87ebcfb2f1baeab184e419908d200db26fa2774137672858533c8384b10828b9",
            "sum": 20114.015098012285,
            "top1": 78
          },
          "net_j": {
            "core": 1.6629554785937994,
            "dram": 0.05517010619666239,
            "package-0": 1.8665523064802017,
            "psys": 2.3650398827265344,
            "uncore": 0.11219752074166243
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.054518764,
          "watts": {
            "core": 34.274199613182724,
            "dram": 1.50575680695916,
            "package-0": 39.96359858781832,
            "psys": 53.53670527086785,
            "uncore": 2.059914637830014
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2530.28332314397,
          "decode_tokens_per_s": 39.39493332412788,
          "end_to_end_ms": 3237.5821020000003,
          "handoff_estimate_ms": 31.659954999999997,
          "handoff_plus_first_decode_ms": 54.518764,
          "output_tokens_per_s_e2e": 4.941959615515566,
          "prefill_tokens_per_s": 2861.920914919376,
          "steady_tpot_ms": 22.858809,
          "steady_tpot_p95_ms": 25.109816749999997,
          "ttft_ms": 2856.822478
        },
        "prefill": {
          "gpu_active_fraction": 0.7632508833922261,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 283,
          "gross_j": {
            "core": 20.284677,
            "dram": 4.12505,
            "package-0": 66.48152,
            "psys": 109.147548,
            "uncore": 35.976532
          },
          "logits": {
            "l2": 655.8555450667305,
            "sha256": "e30c9d66db32ee71e32369110e91308f4656be42fddef82b606d730823ce7e00",
            "sum": 5070.977332231589,
            "top1": 8
          },
          "net_j": {
            "core": 9.509437257980688,
            "dram": 2.7143233854402165,
            "package-0": 50.121297717874114,
            "psys": 80.13247727251171,
            "uncore": 35.97095241021584
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.856822478,
          "watts": {
            "core": 7.10043314073924,
            "dram": 1.4439294117035435,
            "package-0": 23.271141455923534,
            "psys": 38.20592593363094,
            "uncore": 12.59319830932806
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 724.3605232286193,
            "sha256": "546555705e83653d0439120b2a2a1aa452842d59c8f62055b758ef9c1470d280",
            "sum": 84226.24423345178,
            "top1": 42
          },
          "gpu_active_fraction": 0.9696969696969697,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 3.970265,
            "dram": 0.760068,
            "package-0": 7.910442,
            "psys": 12.69882,
            "uncore": 2.317804
          },
          "net_j": {
            "core": 2.7397638025284787,
            "dram": 0.5989671029358344,
            "package-0": 6.042151958104049,
            "psys": 9.3853828858663,
            "uncore": 2.317166826992699
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.32624086,
          "token_latencies_s": [
            0.028226011,
            0.023431866,
            0.023351737,
            0.023308951,
            0.022954834,
            0.022869557,
            0.022782074,
            0.022603696,
            0.022873462,
            0.022848061,
            0.022744495,
            0.022622921,
            0.022841177,
            0.022753609
          ],
          "watts": {
            "core": 12.169735575120786,
            "dram": 2.3297756142501584,
            "package-0": 24.24724481170139,
            "psys": 38.92467669439076,
            "uncore": 7.1045791137259755
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.3119816817343235,
    "compile_target": "NPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396920,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 8192,
      "NPUW_LLM_MIN_RESPONSE_LEN": 8192,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03929801285266876,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 24.575803,
            "dram": 5.850815,
            "package-0": 67.833628,
            "psys": 117.651248,
            "uncore": 28.270435
          },
          "net_j": {
            "core": 13.908018087245983,
            "dram": 4.454156693072982,
            "package-0": 51.63655613811141,
            "psys": 88.92552670192482,
            "uncore": 28.26562753730757
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.047619047619047616,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.50463,
            "dram": 0.196289,
            "package-0": 4.085927,
            "psys": 6.237105,
            "uncore": 0.017395
          },
          "logits": {
            "l2": 1201.221305411654,
            "sha256": "2158e56b7a1abccc4fecc5a5e247427368c958ccb0da6d510335fdeba75d67e2",
            "sum": -118090.05621039867,
            "top1": 151643
          },
          "net_j": {
            "core": 2.712959902789019,
            "dram": 0.0926411830765907,
            "package-0": 2.8839213764985185,
            "psys": 4.105332119454381,
            "uncore": 0.016985060686111747
          },
          "npu_busy_delta_us": 31923,
          "npu_utilization": 0.15209086998550356,
          "seconds": 0.209894256,
          "watts": {
            "core": 16.697122002233353,
            "dram": 0.9351804272337971,
            "package-0": 19.46659750422136,
            "psys": 29.715463009144948,
            "uncore": 0.08287506448008754
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2896.405612149763,
          "decode_tokens_per_s": 26.00807155299033,
          "end_to_end_ms": 2828.333147,
          "handoff_estimate_ms": 186.513786,
          "handoff_plus_first_decode_ms": 209.894256,
          "output_tokens_per_s_e2e": 5.657042211230006,
          "prefill_tokens_per_s": 3631.2131455376903,
          "steady_tpot_ms": 23.38047,
          "steady_tpot_p95_ms": 40.0217218,
          "ttft_ms": 2251.589117
        },
        "prefill": {
          "gpu_active_fraction": 0.9955156950672646,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 223,
          "gross_j": {
            "core": 17.377213,
            "dram": 4.779101,
            "package-0": 57.429541,
            "psys": 100.405321,
            "uncore": 28.25304
          },
          "logits": {
            "l2": 1341.7647947586395,
            "sha256": "3d80fac3fd6fb24da3914052cfc080db8cd8781f8c36dad3eb072ab6b4a68430",
            "sum": 377329.91663436964,
            "top1": 89372
          },
          "net_j": {
            "core": 8.88476633135664,
            "dram": 3.667244482184873,
            "package-0": 44.53532068151385,
            "psys": 77.53725062387568,
            "uncore": 28.24864247662146
          },
          "npu_busy_delta_us": 5500,
          "npu_utilization": 0.0024427192148308826,
          "seconds": 2.251589117,
          "watts": {
            "core": 7.717754926419819,
            "dram": 2.12254578951227,
            "package-0": 25.506226054475995,
            "psys": 44.593092159629585,
            "uncore": 12.548044306433729
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1096.0236381099567,
            "sha256": "c4c493829be4b9bbf7279ba05873d05092edb29b9da7fe7529d2f6f62c3529ec",
            "sum": -78989.97184104472,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 37,
          "gross_j": {
            "core": 3.69396,
            "dram": 0.875425,
            "package-0": 6.31816,
            "psys": 11.008822,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.310291853100323,
            "dram": 0.6942710278115183,
            "package-0": 4.21731408009904,
            "psys": 7.282943958594764,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 333312,
          "npu_utilization": 0.9085789977888877,
          "seconds": 0.366849774,
          "token_latencies_s": [
            0.045486589,
            0.037079101,
            0.02737205,
            0.025429651,
            0.024865887,
            0.024243291,
            0.023558889,
            0.023202051,
            0.022905424,
            0.022743629,
            0.022416924,
            0.022510618,
            0.022520066,
            0.022486937
          ],
          "watts": {
            "core": 10.069407866120152,
            "dram": 2.386331032604098,
            "package-0": 17.22274469767017,
            "psys": 30.009073959522187,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 22.110480000000003,
            "dram": 5.710678000000001,
            "package-0": 64.55726800000001,
            "psys": 111.63399199999999,
            "uncore": 27.856497
          },
          "net_j": {
            "core": 12.046308446961056,
            "dram": 4.393046559466799,
            "package-0": 49.27667214921024,
            "psys": 84.53365303619366,
            "uncore": 27.851962883760265
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.1,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10,
          "gross_j": {
            "core": 2.47674,
            "dram": 0.130676,
            "package-0": 2.85247,
            "psys": 4.055532,
            "uncore": 0.027161
          },
          "logits": {
            "l2": 1219.6496415972515,
            "sha256": "ae511b6f669e4df1ef974305fd17a19af9ab3f992052f05a1989fdf62e185803",
            "sum": -153737.0175356716,
            "top1": 151643
          },
          "net_j": {
            "core": 2.0855684835716124,
            "dram": 0.0794626551190238,
            "package-0": 2.2585479052322626,
            "psys": 3.002203305002304,
            "uncore": 0.02695844519387279
          },
          "npu_busy_delta_us": 29903,
          "npu_utilization": 0.2883309245653194,
          "seconds": 0.103710693,
          "watts": {
            "core": 23.881240481152698,
            "dram": 1.2600050797076439,
            "package-0": 27.504107025878223,
            "psys": 39.10428021149179,
            "uncore": 0.2618919921786657
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3070.121760909111,
          "decode_tokens_per_s": 33.29764299570672,
          "end_to_end_ms": 2668.29808,
          "handoff_estimate_ms": 80.27002750000001,
          "handoff_plus_first_decode_ms": 103.710693,
          "output_tokens_per_s_e2e": 5.996331564275607,
          "prefill_tokens_per_s": 3686.510036786897,
          "steady_tpot_ms": 23.440665499999998,
          "steady_tpot_p95_ms": 31.256373099999998,
          "ttft_ms": 2217.815744
        },
        "prefill": {
          "gpu_active_fraction": 0.9954337899543378,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 219,
          "gross_j": {
            "core": 16.232686,
            "dram": 4.714221,
            "package-0": 55.720072,
            "psys": 97.193721,
            "uncore": 27.829336
          },
          "logits": {
            "l2": 1345.2523121521842,
            "sha256": "1fb23ce8ab84b58eaf4dcab2813fae764591cfe95c15809970ac11972f3a51f4",
            "sum": 380191.34651395306,
            "top1": 89372
          },
          "net_j": {
            "core": 7.8676242822151945,
            "dram": 3.619042100945389,
            "package-0": 43.01926233049613,
            "psys": 74.66866696942994,
            "uncore": 27.825004438566392
          },
          "npu_busy_delta_us": 4158,
          "npu_utilization": 0.0018748176043248434,
          "seconds": 2.217815744,
          "watts": {
            "core": 7.319222096747817,
            "dram": 2.12561436302979,
            "package-0": 25.123850865764258,
            "psys": 43.824073872207116,
            "uncore": 12.548082984480788
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1049.9828109634605,
            "sha256": "241d2b084d70ef6175e411c1167bb1dd4753085dbbef69255281655f2b57032e",
            "sum": -35860.37525684759,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 3.401054,
            "dram": 0.865781,
            "package-0": 5.984726,
            "psys": 10.384739,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.09311568117425,
            "dram": 0.6945418034023864,
            "package-0": 3.998861913481844,
            "psys": 6.862782761761412,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 331642,
          "npu_utilization": 0.9563700109123398,
          "seconds": 0.346771643,
          "token_latencies_s": [
            0.030863468,
            0.031986054,
            0.027159793,
            0.025537363,
            0.02465389,
            0.024103283,
            0.023645351,
            0.02323598,
            0.022970673,
            0.02266734,
            0.022343926,
            0.022241234,
            0.022530057,
            0.02280368
          ],
          "watts": {
            "core": 9.807762741430388,
            "dram": 2.49668915402059,
            "package-0": 17.258406564691334,
            "psys": 29.9469094709108,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 20.748542999999998,
            "dram": 5.682359,
            "package-0": 63.035422000000004,
            "psys": 109.313625,
            "uncore": 27.794362
          },
          "net_j": {
            "core": 10.781093783607577,
            "dram": 4.377390737066231,
            "package-0": 47.90168125081999,
            "psys": 82.4737355023696,
            "uncore": 27.789879570387832
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.1111111111111111,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 9,
          "gross_j": {
            "core": 2.413629,
            "dram": 0.122437,
            "package-0": 2.76574,
            "psys": 3.859488,
            "uncore": 0.025818
          },
          "logits": {
            "l2": 1307.4508778747368,
            "sha256": "70535f068ff0b5d9b427a82892e9a89bfc10f7877cc29c0f02f467344c6b60c9",
            "sum": -241774.7024954781,
            "top1": 151643
          },
          "net_j": {
            "core": 2.0789245420782434,
            "dram": 0.07861649130535636,
            "package-0": 2.2575527613379487,
            "psys": 2.958211206915354,
            "uncore": 0.025644684731027274
          },
          "npu_busy_delta_us": 29919,
          "npu_utilization": 0.33715473586953837,
          "seconds": 0.08873967,
          "watts": {
            "core": 27.198985526991475,
            "dram": 1.3797324240669364,
            "package-0": 31.166895256653532,
            "psys": 43.49225098538229,
            "uncore": 0.29094090613589163
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3099.9136709612935,
          "decode_tokens_per_s": 34.377516319602876,
          "end_to_end_ms": 2642.654238,
          "handoff_estimate_ms": 65.39738550000001,
          "handoff_plus_first_decode_ms": 88.73967,
          "output_tokens_per_s_e2e": 6.054518888596276,
          "prefill_tokens_per_s": 3705.7138493398033,
          "steady_tpot_ms": 23.342284499999998,
          "steady_tpot_p95_ms": 31.46177735,
          "ttft_ms": 2206.322542
        },
        "prefill": {
          "gpu_active_fraction": 0.9908256880733946,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 218,
          "gross_j": {
            "core": 14.744774,
            "dram": 4.695605,
            "package-0": 54.085433,
            "psys": 94.918214,
            "uncore": 27.768544
          },
          "logits": {
            "l2": 1350.4952051518917,
            "sha256": "c479574faa13fa43ac899e354ac4ef981a21beba7736c1d8f95eba41660cb26f",
            "sum": 383524.78882252425,
            "top1": 89372
          },
          "net_j": {
            "core": 6.423061845552484,
            "dram": 3.6061015554311924,
            "package-0": 41.450441679694016,
            "psys": 72.50988968276823,
            "uncore": 27.764234885656805
          },
          "npu_busy_delta_us": 4810,
          "npu_utilization": 0.0021800982895455543,
          "seconds": 2.206322542,
          "watts": {
            "core": 6.682963945350471,
            "dram": 2.128249569414044,
            "package-0": 24.513837832147754,
            "psys": 43.02100540293532,
            "uncore": 12.585895068101967
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 997.4966629813678,
            "sha256": "3397ea1d16a2a94fab376b2782534dfa279ee052f3d3ce5772911a05e328d0af",
            "sum": -29625.88963482529,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 3.59014,
            "dram": 0.864317,
            "package-0": 6.184249,
            "psys": 10.535923,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.2791073959768493,
            "dram": 0.6926726903296824,
            "package-0": 4.1936868097880255,
            "psys": 7.00563461268602,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 332979,
          "npu_utilization": 0.9579592599745083,
          "seconds": 0.347592026,
          "token_latencies_s": [
            0.03278367,
            0.030749989,
            0.027886746,
            0.025532029,
            0.024503872,
            0.02396962,
            0.023495316,
            0.023189253,
            0.022995276,
            0.022671885,
            0.022561967,
            0.022376598,
            0.022445836,
            0.022410346
          ],
          "watts": {
            "core": 10.328602877673609,
            "dram": 2.4865846606043833,
            "package-0": 17.791688351331743,
            "psys": 30.311175780539916,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.3113034395501018,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 396764,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_0_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_0_5b_fp16",
      "size_b": 0.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.03969423286616802,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 22.445866,
            "dram": 4.602283,
            "package-0": 71.283997,
            "psys": 113.860608,
            "uncore": 37.622889
          },
          "net_j": {
            "core": 10.937785561111768,
            "dram": 3.095610689561734,
            "package-0": 53.811090732625104,
            "psys": 82.87217749948371,
            "uncore": 37.61692993369712
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5714285714285714,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2050,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.186091,
            "dram": 0.090271,
            "package-0": 2.560357,
            "psys": 3.409293,
            "uncore": 0.156372
          },
          "logits": {
            "l2": 751.8547055410498,
            "sha256": "87ebcfb2f1baeab184e419908d200db26fa2774137672858533c8384b10828b9",
            "sum": 20114.015098012285,
            "top1": 78
          },
          "net_j": {
            "core": 1.947039034982144,
            "dram": 0.058973601762222695,
            "package-0": 2.1974005017403915,
            "psys": 2.7655848469934394,
            "uncore": 0.15624821510668613
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.063379474,
          "watts": {
            "core": 34.49209755196137,
            "dram": 1.424293928346581,
            "package-0": 40.39725858248681,
            "psys": 53.791752831523965,
            "uncore": 2.4672341080015907
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2684.9162425121453,
          "decode_tokens_per_s": 37.92086452533005,
          "end_to_end_ms": 3051.1193869999997,
          "handoff_estimate_ms": 40.368918,
          "handoff_plus_first_decode_ms": 63.379474,
          "output_tokens_per_s_e2e": 5.243977036156534,
          "prefill_tokens_per_s": 3078.8247076970656,
          "steady_tpot_ms": 23.010556,
          "steady_tpot_p95_ms": 26.6206027,
          "ttft_ms": 2655.558785
        },
        "prefill": {
          "gpu_active_fraction": 0.8015267175572519,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 262,
          "gross_j": {
            "core": 16.263447,
            "dram": 3.74999,
            "package-0": 60.771939,
            "psys": 97.721734,
            "uncore": 35.144258
          },
          "logits": {
            "l2": 655.8555450667305,
            "sha256": "e30c9d66db32ee71e32369110e91308f4656be42fddef82b606d730823ce7e00",
            "sum": 5070.977332231589,
            "top1": 8
          },
          "net_j": {
            "core": 6.247324970103076,
            "dram": 2.438649346747291,
            "package-0": 45.56429752405242,
            "psys": 70.7507805854498,
            "uncore": 35.1390714933892
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.655558785,
          "watts": {
            "core": 6.1243031379551995,
            "dram": 1.4121284082212473,
            "package-0": 22.884802755364348,
            "psys": 36.79893457903625,
            "uncore": 13.234223319970678
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 724.3605232286193,
            "sha256": "546555705e83653d0439120b2a2a1aa452842d59c8f62055b758ef9c1470d280",
            "sum": 84226.24423345178,
            "top1": 42
          },
          "gpu_active_fraction": 0.9696969696969697,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 3.996328,
            "dram": 0.762022,
            "package-0": 7.951701,
            "psys": 12.729581,
            "uncore": 2.322259
          },
          "net_j": {
            "core": 2.743421556026549,
            "dram": 0.5979877410522202,
            "package-0": 6.049392706832295,
            "psys": 9.355812067040477,
            "uncore": 2.321610225201232
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.332181128,
          "token_latencies_s": [
            0.030880964,
            0.024326562,
            0.024064383,
            0.023416013,
            0.023134239,
            0.02300785,
            0.022789099,
            0.022861265,
            0.023013262,
            0.02287143,
            0.022856356,
            0.022959546,
            0.023098143,
            0.022873856
          ],
          "watts": {
            "core": 12.030569057493237,
            "dram": 2.2939954614158573,
            "package-0": 23.937846944754792,
            "psys": 38.3212046892682,
            "uncore": 6.990942001979112
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 26.724359,
            "dram": 4.9849120000000005,
            "package-0": 77.11448899999999,
            "psys": 124.629258,
            "uncore": 38.264795
          },
          "net_j": {
            "core": 14.443491455271161,
            "dram": 3.377064090345446,
            "package-0": 58.46824746296883,
            "psys": 91.55990187090423,
            "uncore": 38.25843577229521
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.5714285714285714,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2000,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.099421,
            "dram": 0.085266,
            "package-0": 2.435175,
            "psys": 3.251029,
            "uncore": 0.127258
          },
          "logits": {
            "l2": 751.8547055410498,
            "sha256": "87ebcfb2f1baeab184e419908d200db26fa2774137672858533c8384b10828b9",
            "sum": 20114.015098012285,
            "top1": 78
          },
          "net_j": {
            "core": 1.874450511994681,
            "dram": 0.05581219084882378,
            "package-0": 2.093598638006628,
            "psys": 2.6452388016049704,
            "uncore": 0.1271415067184893
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.059646074,
          "watts": {
            "core": 35.19797463953788,
            "dram": 1.4295324785332895,
            "package-0": 40.82707941515145,
            "psys": 54.50533089570992,
            "uncore": 2.1335519920389063
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2515.964933093783,
          "decode_tokens_per_s": 38.37590018029254,
          "end_to_end_ms": 3256.007225,
          "handoff_estimate_ms": 36.61781,
          "handoff_plus_first_decode_ms": 59.646074,
          "output_tokens_per_s_e2e": 4.913994009948795,
          "prefill_tokens_per_s": 2853.615810537976,
          "steady_tpot_ms": 23.028264,
          "steady_tpot_p95_ms": 26.27539825,
          "ttft_ms": 2865.136915
        },
        "prefill": {
          "gpu_active_fraction": 0.7667844522968198,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 283,
          "gross_j": {
            "core": 19.898448,
            "dram": 4.136586,
            "package-0": 65.998184,
            "psys": 107.858977,
            "uncore": 35.821686
          },
          "logits": {
            "l2": 655.8555450667305,
            "sha256": "e30c9d66db32ee71e32369110e91308f4656be42fddef82b606d730823ce7e00",
            "sum": 5070.977332231589,
            "top1": 8
          },
          "net_j": {
            "core": 9.091848223297106,
            "dram": 2.721753635905093,
            "package-0": 49.590347265622256,
            "psys": 78.75946141131644,
            "uncore": 35.81609017149324
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.865136915,
          "watts": {
            "core": 6.945025173430499,
            "dram": 1.443765559105925,
            "package-0": 23.03491454613435,
            "psys": 37.6453133654173,
            "uncore": 12.502608797667179
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 724.3605232286193,
            "sha256": "546555705e83653d0439120b2a2a1aa452842d59c8f62055b758ef9c1470d280",
            "sum": 84226.24423345178,
            "top1": 42
          },
          "gpu_active_fraction": 0.9696969696969697,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 4.72649,
            "dram": 0.76306,
            "package-0": 8.68113,
            "psys": 13.519252,
            "uncore": 2.315851
          },
          "net_j": {
            "core": 3.4771927199793744,
            "dram": 0.599498263591529,
            "package-0": 6.784301559339948,
            "psys": 10.155201657982818,
            "uncore": 2.3152040940834846
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.331224236,
          "token_latencies_s": [
            0.030248507,
            0.024136032,
            0.023949366,
            0.023264652,
            0.022959251,
            0.023174113,
            0.023070902,
            0.022985626,
            0.022950056,
            0.022904245,
            0.022951011,
            0.022750485,
            0.023188016,
            0.022661465
          ],
          "watts": {
            "core": 14.269758931529395,
            "dram": 2.303756540327562,
            "package-0": 26.20922340960581,
            "psys": 40.81601081872523,
            "uncore": 6.991792110285069
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 24.925228999999998,
            "dram": 4.999315,
            "package-0": 75.394217,
            "psys": 124.082691,
            "uncore": 38.331445
          },
          "net_j": {
            "core": 12.623916824554648,
            "dram": 3.3887904181444397,
            "package-0": 56.716934046628566,
            "psys": 90.95828250822466,
            "uncore": 38.325075185741646
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.947139,
            "dram": 0.082031,
            "package-0": 2.24853,
            "psys": 3.00219,
            "uncore": 0.1109
          },
          "logits": {
            "l2": 751.8547055410498,
            "sha256": "87ebcfb2f1baeab184e419908d200db26fa2774137672858533c8384b10828b9",
            "sum": 20114.015098012285,
            "top1": 78
          },
          "net_j": {
            "core": 1.7449360244264938,
            "dram": 0.055557981671475125,
            "package-0": 1.9415219235495462,
            "psys": 2.457707113489821,
            "uncore": 0.11079529609699193
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.053609759,
          "watts": {
            "core": 36.32060722377058,
            "dram": 1.5301505086042264,
            "package-0": 41.94255004951617,
            "psys": 56.00081134481504,
            "uncore": 2.068653209203944
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2511.783430078562,
          "decode_tokens_per_s": 38.28302305627987,
          "end_to_end_ms": 3261.427678,
          "handoff_estimate_ms": 30.879824,
          "handoff_plus_first_decode_ms": 53.609759,
          "output_tokens_per_s_e2e": 4.905827011872192,
          "prefill_tokens_per_s": 2849.1685594538003,
          "steady_tpot_ms": 22.729935,
          "steady_tpot_p95_ms": 31.340761199999996,
          "ttft_ms": 2869.6090910000003
        },
        "prefill": {
          "gpu_active_fraction": 0.7640845070422535,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 284,
          "gross_j": {
            "core": 19.043957,
            "dram": 4.153492,
            "package-0": 65.269303,
            "psys": 107.971525,
            "uncore": 35.932159
          },
          "logits": {
            "l2": 655.8555450667305,
            "sha256": "e30c9d66db32ee71e32369110e91308f4656be42fddef82b606d730823ce7e00",
            "sum": 5070.977332231589,
            "top1": 8
          },
          "net_j": {
            "core": 8.220489262848181,
            "dram": 2.7364512320729397,
            "package-0": 48.83585536609726,
            "psys": 78.82658814207586,
            "uncore": 35.92655443699618
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.869609091,
          "watts": {
            "core": 6.636429003423449,
            "dram": 1.447406900482251,
            "package-0": 22.74501541157823,
            "psys": 37.62586525761742,
            "uncore": 12.521621538172077
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 724.3605232286193,
            "sha256": "546555705e83653d0439120b2a2a1aa452842d59c8f62055b758ef9c1470d280",
            "sum": 84226.24423345178,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 34,
          "gross_j": {
            "core": 3.934133,
            "dram": 0.763792,
            "package-0": 7.876384,
            "psys": 13.108976,
            "uncore": 2.288386
          },
          "net_j": {
            "core": 2.6584915372799727,
            "dram": 0.596781204400025,
            "package-0": 5.939556756981758,
            "psys": 9.673987252658968,
            "uncore": 2.2877254526484743
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.338208828,
          "token_latencies_s": [
            0.036353374,
            0.028641662,
            0.02324467,
            0.022730129,
            0.022715643,
            0.022739338,
            0.022724185,
            0.02267181,
            0.022729741,
            0.022777227,
            0.02277903,
            0.022720517,
            0.022663035,
            0.022695453
          ],
          "watts": {
            "core": 11.632259936159915,
            "dram": 2.2583443623180646,
            "package-0": 23.288522793970355,
            "psys": 38.760005401160015,
            "uncore": 6.76619239519082
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.6709308847784996,
    "compile_target": "AUTO:GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393335,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.0455717658624053,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 10.434726999999999,
            "dram": 2.773919,
            "package-0": 26.957389,
            "psys": 44.033579,
            "uncore": 10.756931
          },
          "net_j": {
            "core": 7.956433139067665,
            "dram": 2.2787861220971055,
            "package-0": 22.623376556776588,
            "psys": 34.01359510119007,
            "uncore": 10.754788198706438
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 2.050776,
            "dram": 0.152954,
            "package-0": 2.744622,
            "psys": 3.765798,
            "uncore": 0.375732
          },
          "logits": {
            "l2": 990.1133752307505,
            "sha256": "9ad38a6a770523e4f4bdb5ee760a99cce289de34be1c5c32497a309a9efb2956",
            "sum": 291550.6807441355,
            "top1": 78
          },
          "net_j": {
            "core": 1.9041113370444558,
            "dram": 0.12365218966967964,
            "package-0": 2.4881364862153816,
            "psys": 3.172818456893646,
            "uncore": 0.37560518968325135
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.059704241,
          "watts": {
            "core": 34.348916687509686,
            "dram": 2.5618615602198176,
            "package-0": 45.97030217669127,
            "psys": 63.07421276823535,
            "uncore": 6.293221280545213
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1015.0033191401506,
          "decode_tokens_per_s": 22.221603605698732,
          "end_to_end_ms": 1008.8636960000001,
          "handoff_estimate_ms": 16.121927,
          "handoff_plus_first_decode_ms": 59.704240999999996,
          "output_tokens_per_s_e2e": 15.859426861564852,
          "prefill_tokens_per_s": 3019.3661335044185,
          "steady_tpot_ms": 43.582314,
          "steady_tpot_p95_ms": 45.64459405,
          "ttft_ms": 333.84490500000004
        },
        "prefill": {
          "gpu_active_fraction": 0.8181818181818182,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 1.807735,
            "dram": 0.684874,
            "package-0": 8.468484,
            "psys": 13.987269,
            "uncore": 5.107287
          },
          "logits": {
            "l2": 1146.1615279083735,
            "sha256": "9992a7f0f85fa0507bed1eb81bdf57b5726e3f270c5327fd51897bcdd6e10001",
            "sum": 371186.4616530363,
            "top1": 291
          },
          "net_j": {
            "core": 0.9876383091928818,
            "dram": 0.5210286864643364,
            "package-0": 7.034308126575828,
            "psys": 10.671538058285421,
            "uncore": 5.10657792175968
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.333844905,
          "watts": {
            "core": 5.41489467991132,
            "dram": 2.0514735727358184,
            "package-0": 25.36652161877384,
            "psys": 41.89750626866688,
            "uncore": 15.298382343142244
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 946.9425538034262,
            "sha256": "d666fa852f546efdb063aa00cdee49c7db96fa7b0654c593aeeadfd6ccfa0e85",
            "sum": 270664.63174004806,
            "top1": 42
          },
          "gpu_active_fraction": 0.9836065573770492,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 6.576216,
            "dram": 1.936091,
            "package-0": 15.744283,
            "psys": 26.280512,
            "uncore": 5.273912
          },
          "net_j": {
            "core": 5.064683492830327,
            "dram": 1.6341052459630896,
            "package-0": 13.100931943985378,
            "psys": 20.169238586011005,
            "uncore": 5.272605087263507
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.61531455,
          "token_latencies_s": [
            0.047299509,
            0.044753486,
            0.044034376,
            0.043584331,
            0.044182667,
            0.043532346,
            0.043443539,
            0.043244585,
            0.043237435,
            0.043233129,
            0.043337202,
            0.043677565,
            0.043580297,
            0.044111402
          ],
          "watts": {
            "core": 10.687567846396611,
            "dram": 3.146506124387925,
            "package-0": 25.587373157355046,
            "psys": 42.710694879553884,
            "uncore": 8.571082871354172
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 9.975317,
            "dram": 2.820732,
            "package-0": 26.794425999999998,
            "psys": 44.762031,
            "uncore": 10.913119
          },
          "net_j": {
            "core": 7.410699591157208,
            "dram": 2.308352730383246,
            "package-0": 22.30945190663298,
            "psys": 34.39303257398403,
            "uncore": 10.91090156098367
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.512021,
            "dram": 0.145202,
            "package-0": 2.110224,
            "psys": 3.011894,
            "uncore": 0.301147
          },
          "logits": {
            "l2": 990.1133752307505,
            "sha256": "9ad38a6a770523e4f4bdb5ee760a99cce289de34be1c5c32497a309a9efb2956",
            "sum": 291550.6807441355,
            "top1": 78
          },
          "net_j": {
            "core": 1.3820457006522604,
            "dram": 0.11923452090608569,
            "package-0": 1.8829246576722616,
            "psys": 2.4863911824215355,
            "uncore": 0.30103461976779106
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.052910336,
          "watts": {
            "core": 28.577044001383776,
            "dram": 2.7443031168806034,
            "package-0": 39.88302020988867,
            "psys": 56.92449203119783,
            "uncore": 5.691647847407357
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 980.8388908137426,
          "decode_tokens_per_s": 22.553990554533097,
          "end_to_end_ms": 1044.00428,
          "handoff_estimate_ms": 9.532815,
          "handoff_plus_first_decode_ms": 52.910336,
          "output_tokens_per_s_e2e": 15.325607668964729,
          "prefill_tokens_per_s": 2660.0973240940975,
          "steady_tpot_ms": 43.377521,
          "steady_tpot_p95_ms": 45.06358405,
          "ttft_ms": 378.93350399999997
        },
        "prefill": {
          "gpu_active_fraction": 0.7105263157894737,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 38,
          "gross_j": {
            "core": 2.321222,
            "dram": 0.744322,
            "package-0": 9.391822,
            "psys": 15.725607,
            "uncore": 5.351488
          },
          "logits": {
            "l2": 1146.1615279083735,
            "sha256": "9992a7f0f85fa0507bed1eb81bdf57b5726e3f270c5327fd51897bcdd6e10001",
            "sum": 371186.4616530363,
            "top1": 291
          },
          "net_j": {
            "core": 1.3903642633322923,
            "dram": 0.5583479808007326,
            "package-0": 7.7639484547413655,
            "psys": 11.9620583423827,
            "uncore": 5.350683154701507
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.378933504,
          "watts": {
            "core": 6.125671062329713,
            "dram": 1.9642549210956022,
            "package-0": 24.78488151842071,
            "psys": 41.499647917118466,
            "uncore": 14.122498917382613
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 946.9425538034262,
            "sha256": "d666fa852f546efdb063aa00cdee49c7db96fa7b0654c593aeeadfd6ccfa0e85",
            "sum": 270664.63174004806,
            "top1": 42
          },
          "gpu_active_fraction": 0.9836065573770492,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 6.142074,
            "dram": 1.931208,
            "package-0": 15.29238,
            "psys": 26.02453,
            "uncore": 5.260484
          },
          "net_j": {
            "core": 4.6382896271726555,
            "dram": 1.6307702286764276,
            "package-0": 12.662578794219353,
            "psys": 19.944583049179794,
            "uncore": 5.259183786514372
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.61216044,
          "token_latencies_s": [
            0.04690371,
            0.044072747,
            0.043788318,
            0.043391651,
            0.043325283,
            0.043351694,
            0.04308949,
            0.043349156,
            0.043239634,
            0.043385532,
            0.04336951,
            0.043548176,
            0.043253087,
            0.044067795
          ],
          "watts": {
            "core": 10.033438292745608,
            "dram": 3.1547415902928977,
            "package-0": 24.981000078998896,
            "psys": 42.51259686104512,
            "uncore": 8.5933092964975
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 9.011085000000001,
            "dram": 2.8106,
            "package-0": 25.564327,
            "psys": 43.079784000000004,
            "uncore": 10.739536000000001
          },
          "net_j": {
            "core": 6.486228325435229,
            "dram": 2.3061644400058015,
            "package-0": 21.148886031751488,
            "psys": 32.87154210231628,
            "uncore": 10.737352939212252
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 1.789241,
            "dram": 0.147949,
            "package-0": 2.376154,
            "psys": 3.312919,
            "uncore": 0.277831
          },
          "logits": {
            "l2": 990.1133752307505,
            "sha256": "9ad38a6a770523e4f4bdb5ee760a99cce289de34be1c5c32497a309a9efb2956",
            "sum": 291550.6807441355,
            "top1": 78
          },
          "net_j": {
            "core": 1.6528156578115332,
            "dram": 0.12069288103788024,
            "package-0": 2.137574895263247,
            "psys": 2.76133802985475,
            "uncore": 0.27771304288267645
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.055536019,
          "watts": {
            "core": 32.21766760055308,
            "dram": 2.6640188235314453,
            "package-0": 42.785817975177515,
            "psys": 59.65351963740865,
            "uncore": 5.002717245541133
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 996.2848663814116,
          "decode_tokens_per_s": 22.477790538086463,
          "end_to_end_ms": 1027.8184829999998,
          "handoff_estimate_ms": 12.014416,
          "handoff_plus_first_decode_ms": 55.536018999999996,
          "output_tokens_per_s_e2e": 15.566951037209556,
          "prefill_tokens_per_s": 2796.169961729199,
          "steady_tpot_ms": 43.521603,
          "steady_tpot_p95_ms": 44.418626749999994,
          "ttft_ms": 360.493108
        },
        "prefill": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 2.164362,
            "dram": 0.730161,
            "package-0": 9.003578,
            "psys": 15.070762,
            "uncore": 5.185656
          },
          "logits": {
            "l2": 1146.1615279083735,
            "sha256": "9992a7f0f85fa0507bed1eb81bdf57b5726e3f270c5327fd51897bcdd6e10001",
            "sum": 371186.4616530363,
            "top1": 291
          },
          "net_j": {
            "core": 1.2788034651780502,
            "dram": 0.5532372086732884,
            "package-0": 7.454923189679338,
            "psys": 11.490362435506256,
            "uncore": 5.184890321642055
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.360493108,
          "watts": {
            "core": 6.003892867766004,
            "dram": 2.0254506502243586,
            "package-0": 24.97572852349787,
            "psys": 41.8059642904463,
            "uncore": 14.384896368115863
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 946.9425538034262,
            "sha256": "d666fa852f546efdb063aa00cdee49c7db96fa7b0654c593aeeadfd6ccfa0e85",
            "sum": 270664.63174004806,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 5.057482,
            "dram": 1.93249,
            "package-0": 14.184595,
            "psys": 24.696103,
            "uncore": 5.276049
          },
          "net_j": {
            "core": 3.5546092024456453,
            "dram": 1.6322343502946326,
            "package-0": 11.556387946808904,
            "psys": 18.61984163695528,
            "uncore": 5.274749574687521
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.611789356,
          "token_latencies_s": [
            0.04513961,
            0.044030405,
            0.043728651,
            0.043572165,
            0.043488254,
            0.043425616,
            0.043927817,
            0.043484576,
            0.043170101,
            0.043554952,
            0.043998679,
            0.043355741,
            0.043409205,
            0.043478174
          ],
          "watts": {
            "core": 8.266704790463862,
            "dram": 3.1587506076192673,
            "package-0": 23.18542299058894,
            "psys": 40.367003377548144,
            "uncore": 8.623963376048016
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.4886896423995495,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393404,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04515166487544775,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.611725,
            "dram": 3.028618,
            "package-0": 25.76147,
            "psys": 43.123241,
            "uncore": 5.447313
          },
          "net_j": {
            "core": 8.524323082877697,
            "dram": 2.411792763668828,
            "package-0": 20.36225623565407,
            "psys": 30.640573796409228,
            "uncore": 5.446590373366908
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 23,
          "gross_j": {
            "core": 3.662466,
            "dram": 0.273559,
            "package-0": 4.408558,
            "psys": 6.869672,
            "uncore": 0.0
          },
          "logits": {
            "l2": 991.8733295788431,
            "sha256": "8ff682a14bc39705e7bd3c784b3399bc49fa56c100079d57fa5ecae59ced5fc7",
            "sum": 291802.32683754154,
            "top1": 78
          },
          "net_j": {
            "core": 3.087492529951442,
            "dram": 0.15868631543633147,
            "package-0": 3.40305085938338,
            "psys": 4.544998157664751,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 65246,
          "npu_utilization": 0.27875739178014153,
          "seconds": 0.234060161,
          "watts": {
            "core": 15.647541146483277,
            "dram": 1.1687550706247698,
            "package-0": 18.835148968388516,
            "psys": 29.350026807851336,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 814.7551119601545,
          "decode_tokens_per_s": 16.364903911732,
          "end_to_end_ms": 1256.8193620000002,
          "handoff_estimate_ms": 186.644851,
          "handoff_plus_first_decode_ms": 234.060161,
          "output_tokens_per_s_e2e": 12.730548624377414,
          "prefill_tokens_per_s": 2962.756597647449,
          "steady_tpot_ms": 47.415310000000005,
          "steady_tpot_p95_ms": 54.75503085,
          "ttft_ms": 340.223696
        },
        "prefill": {
          "gpu_active_fraction": 0.9705882352941176,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 34,
          "gross_j": {
            "core": 2.118464,
            "dram": 0.71826,
            "package-0": 9.260231,
            "psys": 15.358969,
            "uncore": 5.447313
          },
          "logits": {
            "l2": 1145.1161856425492,
            "sha256": "2ae057c3aedd79f7ab7ca960943e32321cda2cc25585ad5dfc0e9014c0ed5485",
            "sum": 370676.0797875095,
            "top1": 291
          },
          "net_j": {
            "core": 1.2826976814342963,
            "dram": 0.5512840860096928,
            "package-0": 7.798652257708277,
            "psys": 11.979884224296459,
            "uncore": 5.446590373366908
          },
          "npu_busy_delta_us": 8568,
          "npu_utilization": 0.025183431080003312,
          "seconds": 0.340223696,
          "watts": {
            "core": 6.226679754839886,
            "dram": 2.111140430383191,
            "package-0": 27.218065963283166,
            "psys": 45.1437368430681,
            "uncore": 16.01097473234198
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 954.2613549585964,
            "sha256": "1873f2d4527206e819430b61339692e2b6ffc74aac45fa6cf5359f3882dc15ad",
            "sum": 273881.8022980206,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 67,
          "gross_j": {
            "core": 5.830795,
            "dram": 2.036799,
            "package-0": 12.092681,
            "psys": 20.8946,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.154132871491957,
            "dram": 1.7018223622228037,
            "package-0": 9.160553118562412,
            "psys": 14.115691414448019,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 669584,
          "npu_utilization": 0.9810244230444831,
          "seconds": 0.682535505,
          "token_latencies_s": [
            0.05903209,
            0.052451999,
            0.049565249,
            0.04833845,
            0.04773908,
            0.047415147,
            0.0474718,
            0.047261906,
            0.047311857,
            0.047415473,
            0.047000139,
            0.047188932,
            0.046964112,
            0.047347732
          ],
          "watts": {
            "core": 8.54284496159654,
            "dram": 2.9841656369217007,
            "package-0": 17.717292230826878,
            "psys": 30.613205975270105,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.833534,
            "dram": 2.916252,
            "package-0": 22.609560000000002,
            "psys": 38.069848,
            "uncore": 5.403734
          },
          "net_j": {
            "core": 6.1429467238449185,
            "dram": 2.3787054877432166,
            "package-0": 17.904291218820102,
            "psys": 27.19154110895417,
            "uncore": 5.402855637825849
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.14285714285714285,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.207514,
            "dram": 0.164245,
            "package-0": 2.647088,
            "psys": 3.674551,
            "uncore": 0.023133
          },
          "logits": {
            "l2": 991.8733295788431,
            "sha256": "8ff682a14bc39705e7bd3c784b3399bc49fa56c100079d57fa5ecae59ced5fc7",
            "sum": 291802.32683754154,
            "top1": 78
          },
          "net_j": {
            "core": 2.0388429292134433,
            "dram": 0.13054657786611612,
            "package-0": 2.352117927594727,
            "psys": 2.992597402761729,
            "uncore": 0.022987162335481846
          },
          "npu_busy_delta_us": 61351,
          "npu_utilization": 0.8935139777970567,
          "seconds": 0.068662608,
          "watts": {
            "core": 32.15016242901814,
            "dram": 2.3920588626636494,
            "package-0": 38.55210393406554,
            "psys": 53.516041802548486,
            "uncore": 0.3369082630825791
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 934.9172639534885,
          "decode_tokens_per_s": 19.98933057486968,
          "end_to_end_ms": 1095.2840849999998,
          "handoff_estimate_ms": 21.257724000000007,
          "handoff_plus_first_decode_ms": 68.662608,
          "output_tokens_per_s_e2e": 14.608082249273258,
          "prefill_tokens_per_s": 2922.72380879346,
          "steady_tpot_ms": 47.404883999999996,
          "steady_tpot_p95_ms": 54.15565405,
          "ttft_ms": 344.883768
        },
        "prefill": {
          "gpu_active_fraction": 0.9714285714285714,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 1.550472,
            "dram": 0.718504,
            "package-0": 8.638405,
            "psys": 14.633446,
            "uncore": 5.380601
          },
          "logits": {
            "l2": 1145.1161856425492,
            "sha256": "2ae057c3aedd79f7ab7ca960943e32321cda2cc25585ad5dfc0e9014c0ed5485",
            "sum": 370676.0797875095,
            "top1": 291
          },
          "net_j": {
            "core": 0.7032581212981822,
            "dram": 0.549241003156238,
            "package-0": 7.1568068932347675,
            "psys": 11.208077621672608,
            "uncore": 5.379868475490367
          },
          "npu_busy_delta_us": 6651,
          "npu_utilization": 0.019284757988378278,
          "seconds": 0.344883768,
          "watts": {
            "core": 4.495636338559141,
            "dram": 2.0833221701521194,
            "package-0": 25.0472936145838,
            "psys": 42.43008038580697,
            "uncore": 15.601201039997918
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 954.2613549585964,
            "sha256": "1873f2d4527206e819430b61339692e2b6ffc74aac45fa6cf5359f3882dc15ad",
            "sum": 273881.8022980206,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 5.075548,
            "dram": 2.033503,
            "package-0": 11.324067,
            "psys": 19.761851,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.4008456733332926,
            "dram": 1.6989179067208622,
            "package-0": 8.395366397990607,
            "psys": 12.990866084519833,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 667292,
          "npu_utilization": 0.9788104591996393,
          "seconds": 0.681737709,
          "token_latencies_s": [
            0.058657777,
            0.051731434,
            0.049256804,
            0.048283387,
            0.0475006,
            0.047398644,
            0.047336664,
            0.047613775,
            0.04734079,
            0.047295518,
            0.047350557,
            0.047411124,
            0.047344449,
            0.04718353
          ],
          "watts": {
            "core": 7.445015778054901,
            "dram": 2.982823119734455,
            "package-0": 16.61059209503108,
            "psys": 28.987469431590444,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.460244,
            "dram": 2.9150310000000004,
            "package-0": 22.19733,
            "psys": 37.127041,
            "uncore": 5.402574
          },
          "net_j": {
            "core": 5.779125186174984,
            "dram": 2.3793761710002768,
            "package-0": 17.50861955929925,
            "psys": 26.287016026038074,
            "uncore": 5.401700764583309
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.14285714285714285,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.097712,
            "dram": 0.160644,
            "package-0": 2.520501,
            "psys": 3.504813,
            "uncore": 0.020081
          },
          "logits": {
            "l2": 991.8733295788431,
            "sha256": "8ff682a14bc39705e7bd3c784b3399bc49fa56c100079d57fa5ecae59ced5fc7",
            "sum": 291802.32683754154,
            "top1": 78
          },
          "net_j": {
            "core": 1.9349193963468012,
            "dram": 0.12812002388128813,
            "package-0": 2.2358111248323014,
            "psys": 2.846626618972228,
            "uncore": 0.019940245020459643
          },
          "npu_busy_delta_us": 62070,
          "npu_utilization": 0.9366285314343672,
          "seconds": 0.066269602,
          "watts": {
            "core": 31.654211534271777,
            "dram": 2.4240978540960607,
            "package-0": 38.03404462878772,
            "psys": 52.887189514130476,
            "uncore": 0.30301977669942853
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 938.2189560865737,
          "decode_tokens_per_s": 20.091977160201182,
          "end_to_end_ms": 1091.429664,
          "handoff_estimate_ms": 18.838401499999996,
          "handoff_plus_first_decode_ms": 66.26960199999999,
          "output_tokens_per_s_e2e": 14.659671188852714,
          "prefill_tokens_per_s": 2922.8996656289773,
          "steady_tpot_ms": 47.4312005,
          "steady_tpot_p95_ms": 53.66817335,
          "ttft_ms": 344.863018
        },
        "prefill": {
          "gpu_active_fraction": 0.9714285714285714,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 1.836543,
            "dram": 0.719053,
            "package-0": 8.928688,
            "psys": 14.844322,
            "uncore": 5.382493
          },
          "logits": {
            "l2": 1145.1161856425492,
            "sha256": "2ae057c3aedd79f7ab7ca960943e32321cda2cc25585ad5dfc0e9014c0ed5485",
            "sum": 370676.0797875095,
            "top1": 291
          },
          "net_j": {
            "core": 0.989380094088479,
            "dram": 0.5498001868978646,
            "package-0": 7.447179033877405,
            "psys": 11.419159709637437,
            "uncore": 5.381760519562849
          },
          "npu_busy_delta_us": 6709,
          "npu_utilization": 0.01945410104831826,
          "seconds": 0.344863018,
          "watts": {
            "core": 5.325427500608372,
            "dram": 2.085039457608644,
            "package-0": 25.890534890580813,
            "psys": 43.044110922905624,
            "uncore": 15.607625982093563
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 954.2613549585964,
            "sha256": "1873f2d4527206e819430b61339692e2b6ffc74aac45fa6cf5359f3882dc15ad",
            "sum": 273881.8022980206,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 4.525989,
            "dram": 2.035334,
            "package-0": 10.748141,
            "psys": 18.777906,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.854825695739704,
            "dram": 1.701455960221124,
            "package-0": 7.825629400589544,
            "psys": 12.02122969742841,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 667254,
          "npu_utilization": 0.9808274280844883,
          "seconds": 0.680297044,
          "token_latencies_s": [
            0.058306184,
            0.051170783,
            0.049181799,
            0.048039222,
            0.047425901,
            0.047389814,
            0.047449361,
            0.047272791,
            0.047186459,
            0.047390032,
            0.047141211,
            0.047494376,
            0.0474365,
            0.047380614
          ],
          "watts": {
            "core": 6.6529599678813245,
            "dram": 2.991831315380521,
            "package-0": 15.799188155813889,
            "psys": 27.6025100588266,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.697680806741118,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393554,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU+GPU",
      "key": "npu_gpu_combined",
      "label": "NPU + iGPU (subgraph combination)",
      "mode": "combined",
      "prefill_device": "NPU+GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.048760765232145786,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 57.393957,
            "dram": 9.769201,
            "package-0": 121.404231,
            "psys": 215.112425,
            "uncore": 33.586767
          },
          "net_j": {
            "core": 41.64916887460104,
            "dram": 6.623584421033316,
            "package-0": 93.869922527911,
            "psys": 151.45471031425197,
            "uncore": 33.57315362150045
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.422680412371134,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 97,
          "gross_j": {
            "core": 13.39169,
            "dram": 1.015806,
            "package-0": 18.551955,
            "psys": 29.517502,
            "uncore": 2.178339
          },
          "logits": {
            "l2": 1882.7353047579231,
            "sha256": "1902aae974c485506c3338d22a40ea0aea7c444494aded21d1f363e35a2038ca",
            "sum": 661811.1864299811,
            "top1": 69754
          },
          "net_j": {
            "core": 10.9813444259671,
            "dram": 0.5342483608415665,
            "package-0": 14.336769757037285,
            "psys": 19.77223965516588,
            "uncore": 2.176254948671225
          },
          "npu_busy_delta_us": 8736,
          "npu_utilization": 0.008903353559002255,
          "seconds": 0.981203312,
          "watts": {
            "core": 13.648231550200883,
            "dram": 1.0352655637998907,
            "package-0": 18.90735056956269,
            "psys": 30.08296205180359,
            "uncore": 2.2200689432650424
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 159.76566179338394,
          "decode_tokens_per_s": 2.512044668460185,
          "end_to_end_ms": 6409.387277,
          "handoff_estimate_ms": 626.926127,
          "handoff_plus_first_decode_ms": 981.203312,
          "output_tokens_per_s_e2e": 2.496338465521624,
          "prefill_tokens_per_s": 2300.551129431613,
          "steady_tpot_ms": 354.277185,
          "steady_tpot_p95_ms": 366.59889565000003,
          "ttft_ms": 438.155878
        },
        "prefill": {
          "gpu_active_fraction": 0.06818181818181818,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 44,
          "gross_j": {
            "core": 1.441524,
            "dram": 0.645018,
            "package-0": 7.825969,
            "psys": 13.256497,
            "uncore": 0.161193
          },
          "logits": {
            "l2": 1145.1872934436176,
            "sha256": "241c43a634ad9ad84cb70858b42518580d814415f305684145ca096fb8068099",
            "sum": 370344.48207614943,
            "top1": 291
          },
          "net_j": {
            "core": 0.36518531630648243,
            "dram": 0.4299786624716829,
            "package-0": 5.94368001006743,
            "psys": 8.904754678281122,
            "uncore": 0.16026236785823192
          },
          "npu_busy_delta_us": 393312,
          "npu_utilization": 0.8976531406934588,
          "seconds": 0.438155878,
          "watts": {
            "core": 3.289979827681326,
            "dram": 1.4721199289719444,
            "package-0": 17.861152601038484,
            "psys": 30.255207485770622,
            "uncore": 0.36788962123657737
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 3230.8548509007255,
            "sha256": "3dc7ccdc989a4822f462f6221f23dadd86c14033eb527cada2f58c1e598e55d5",
            "sum": 1207703.5942871496,
            "top1": 283
          },
          "gpu_active_fraction": 0.9979716024340771,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 493,
          "gross_j": {
            "core": 42.560743,
            "dram": 8.108377,
            "package-0": 95.026307,
            "psys": 172.338426,
            "uncore": 31.247235
          },
          "net_j": {
            "core": 30.302639132327453,
            "dram": 5.659357397720067,
            "package-0": 73.58947276080629,
            "psys": 122.77771598080497,
            "uncore": 31.236636304970997
          },
          "npu_busy_delta_us": 124876,
          "npu_utilization": 0.025025109643235562,
          "seconds": 4.990028087,
          "token_latencies_s": [
            0.369897229,
            0.353039063,
            0.351452315,
            0.352740454,
            0.351920798,
            0.353892769,
            0.354661601,
            0.352234731,
            0.352330242,
            0.357071229,
            0.357833043,
            0.355913636,
            0.362166688,
            0.36482287
          ],
          "watts": {
            "core": 8.529159006314828,
            "dram": 1.6249161044050855,
            "package-0": 19.043240908315152,
            "psys": 34.536564322949474,
            "uncore": 6.261935695593611
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 47.142763,
            "dram": 10.038181999999999,
            "package-0": 111.652301,
            "psys": 202.13907,
            "uncore": 34.269138
          },
          "net_j": {
            "core": 32.31522683642132,
            "dram": 7.027489271528754,
            "package-0": 85.2990121552312,
            "psys": 141.21180378549022,
            "uncore": 34.256108535464314
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.975609756097561,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 41,
          "gross_j": {
            "core": 5.00249,
            "dram": 0.647276,
            "package-0": 8.948097,
            "psys": 15.133933,
            "uncore": 2.271478
          },
          "logits": {
            "l2": 3233.3883385391673,
            "sha256": "d7d4781f681a7ddbc3a5ad43821a5de1a0bc4a68985453853a781c9d9377abee",
            "sum": 1208728.0883883238,
            "top1": 283
          },
          "net_j": {
            "core": 3.9943947589986752,
            "dram": 0.4458708697425975,
            "package-0": 7.185151375460387,
            "psys": 11.058105575926831,
            "uncore": 2.2706063730228667
          },
          "npu_busy_delta_us": 11230,
          "npu_utilization": 0.027365191936897616,
          "seconds": 0.410375342,
          "watts": {
            "core": 12.190035530935969,
            "dram": 1.5772780032188192,
            "package-0": 21.80466534950826,
            "psys": 36.87827081969267,
            "uncore": 5.535123014286761
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 166.9255416649712,
          "decode_tokens_per_s": 2.64201365027035,
          "end_to_end_ms": 6134.471632000001,
          "handoff_estimate_ms": 37.32217599999993,
          "handoff_plus_first_decode_ms": 410.375342,
          "output_tokens_per_s_e2e": 2.608211588515175,
          "prefill_tokens_per_s": 2205.766827255172,
          "steady_tpot_ms": 373.05316600000003,
          "steady_tpot_p95_ms": 389.02324304999996,
          "ttft_ms": 456.98393300000004
        },
        "prefill": {
          "gpu_active_fraction": 0.06521739130434782,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1950,
          "gpu_sample_count": 46,
          "gross_j": {
            "core": 0.880674,
            "dram": 0.635862,
            "package-0": 7.240643,
            "psys": 12.296905,
            "uncore": 0.160461
          },
          "logits": {
            "l2": 1145.1872934436176,
            "sha256": "241c43a634ad9ad84cb70858b42518580d814415f305684145ca096fb8068099",
            "sum": 370344.48207614943,
            "top1": 291
          },
          "net_j": {
            "core": 0.0,
            "dram": 0.41158217825718907,
            "package-0": 5.2774699174241615,
            "psys": 7.758163375308972,
            "uncore": 0.1594903775396495
          },
          "npu_busy_delta_us": 399276,
          "npu_utilization": 0.8737199957531111,
          "seconds": 0.456983933,
          "watts": {
            "core": 1.9271443401052788,
            "dram": 1.3914318515001272,
            "package-0": 15.844414818846596,
            "psys": 26.908834451297878,
            "uncore": 0.35113050681368263
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 3244.1967264885366,
            "sha256": "1c7b735d072ce209290e8ce01671cb97d65a146d735d72abae9c93f26e69c07c",
            "sum": 1213023.6846235134,
            "top1": 283
          },
          "gpu_active_fraction": 0.9980769230769231,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 520,
          "gross_j": {
            "core": 41.259599,
            "dram": 8.755044,
            "package-0": 95.463561,
            "psys": 174.708232,
            "uncore": 31.837199
          },
          "net_j": {
            "core": 28.320832077422644,
            "dram": 6.170036223528967,
            "package-0": 72.83639086234665,
            "psys": 122.39553483425442,
            "uncore": 31.826011784901794
          },
          "npu_busy_delta_us": 127020,
          "npu_utilization": 0.024115680735610323,
          "seconds": 5.267112357,
          "token_latencies_s": [
            0.41296307,
            0.372307188,
            0.371511816,
            0.376132567,
            0.372543283,
            0.374152864,
            0.375724969,
            0.372111054,
            0.373419747,
            0.372686585,
            0.371189293,
            0.372352259,
            0.374926239,
            0.375027675
          ],
          "watts": {
            "core": 7.833438173227107,
            "dram": 1.6622094625273245,
            "package-0": 18.124458817197773,
            "psys": 33.16964214135522,
            "uncore": 6.044526268304931
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 50.119991,
            "dram": 10.028783,
            "package-0": 115.11945599999999,
            "psys": 207.57100200000002,
            "uncore": 34.613558999999995
          },
          "net_j": {
            "core": 35.25964192780879,
            "dram": 7.013005840000079,
            "package-0": 88.72166195168798,
            "psys": 146.5408423515746,
            "uncore": 34.60050753141864
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9767441860465116,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 43,
          "gross_j": {
            "core": 5.127551,
            "dram": 0.669066,
            "package-0": 9.258339,
            "psys": 15.566061,
            "uncore": 2.369745
          },
          "logits": {
            "l2": 3244.8463463137473,
            "sha256": "988d0c153e14ceb46d50b742f501554610cb77a788fc4aae049ed148b7abb1c7",
            "sum": 1213281.8226122651,
            "top1": 283
          },
          "net_j": {
            "core": 4.053848076303634,
            "dram": 0.4545532551624949,
            "package-0": 7.380659397519452,
            "psys": 11.224975313133395,
            "uncore": 2.368816646809094
          },
          "npu_busy_delta_us": 7291,
          "npu_utilization": 0.016681045593745274,
          "seconds": 0.437082913,
          "watts": {
            "core": 11.731300509567163,
            "dram": 1.5307530450177997,
            "package-0": 21.18211150477987,
            "psys": 35.61351985406942,
            "uncore": 5.421728760190632
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 166.64411454288808,
          "decode_tokens_per_s": 2.641463475621028,
          "end_to_end_ms": 6144.8314740000005,
          "handoff_estimate_ms": 65.30061900000005,
          "handoff_plus_first_decode_ms": 437.082913,
          "output_tokens_per_s_e2e": 2.603814289732626,
          "prefill_tokens_per_s": 2162.3419166902218,
          "steady_tpot_ms": 371.782294,
          "steady_tpot_p95_ms": 390.4651399,
          "ttft_ms": 466.161245
        },
        "prefill": {
          "gpu_active_fraction": 0.06521739130434782,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 1900,
          "gpu_sample_count": 46,
          "gross_j": {
            "core": 0.910582,
            "dram": 0.657713,
            "package-0": 7.353313,
            "psys": 12.34909,
            "uncore": 0.1604
          },
          "logits": {
            "l2": 1145.1872934436176,
            "sha256": "241c43a634ad9ad84cb70858b42518580d814415f305684145ca096fb8068099",
            "sum": 370344.48207614943,
            "top1": 291
          },
          "net_j": {
            "core": 0.0,
            "dram": 0.42892911202902656,
            "package-0": 5.350714785405788,
            "psys": 7.719199770359711,
            "uncore": 0.15940988515816823
          },
          "npu_busy_delta_us": 399904,
          "npu_utilization": 0.8578662518373873,
          "seconds": 0.466161245,
          "watts": {
            "core": 1.9533627253805708,
            "dram": 1.4109130843770592,
            "package-0": 15.77418345877294,
            "psys": 26.491026726170684,
            "uncore": 0.3440869478542773
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 3249.8824613058905,
            "sha256": "2cbd4ab13a8143a7e97bc1e3d8bae02c5d101b7bab2ad7c8989726a89a5ff0ea",
            "sum": 1215262.375621643,
            "top1": 283
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 515,
          "gross_j": {
            "core": 44.081858,
            "dram": 8.702004,
            "package-0": 98.507804,
            "psys": 179.655851,
            "uncore": 32.083414
          },
          "net_j": {
            "core": 31.205793851505156,
            "dram": 6.1295234728085575,
            "package-0": 75.99028776876274,
            "psys": 127.59666726808152,
            "uncore": 32.07228099945138
          },
          "npu_busy_delta_us": 126205,
          "npu_utilization": 0.024077629998599456,
          "seconds": 5.241587316,
          "token_latencies_s": [
            0.414582814,
            0.3774787,
            0.366906066,
            0.368730497,
            0.370806839,
            0.372245928,
            0.371846289,
            0.371718299,
            0.371105392,
            0.372687579,
            0.370395877,
            0.37225142,
            0.373026945,
            0.367747079
          ],
          "watts": {
            "core": 8.410020732734846,
            "dram": 1.6601848782404218,
            "package-0": 18.793506253211483,
            "psys": 34.2750850399074,
            "uncore": 6.120934759984831
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 4.743245515041053,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393251,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05435913521796465,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 9.007484,
            "dram": 2.775261,
            "package-0": 25.480098,
            "psys": 42.233107000000004,
            "uncore": 10.7233
          },
          "net_j": {
            "core": 6.520063238649946,
            "dram": 2.2783046787430323,
            "package-0": 21.130124535929944,
            "psys": 32.176222152245636,
            "uncore": 10.72114930733636
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.052119,
            "dram": 0.154296,
            "package-0": 2.766167,
            "psys": 3.78179,
            "uncore": 0.391722
          },
          "logits": {
            "l2": 990.1133752307505,
            "sha256": "9ad38a6a770523e4f4bdb5ee760a99cce289de34be1c5c32497a309a9efb2956",
            "sum": 291550.6807441355,
            "top1": 78
          },
          "net_j": {
            "core": 1.900753122736674,
            "dram": 0.1240549444064184,
            "package-0": 2.5014600554845963,
            "psys": 3.1698029887271777,
            "uncore": 0.39159112488356845
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.061618011,
          "watts": {
            "core": 33.30388252876257,
            "dram": 2.5040730379953353,
            "package-0": 44.89218257953831,
            "psys": 61.37474966532107,
            "uncore": 6.357264599144559
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1011.2790460451487,
          "decode_tokens_per_s": 22.150324892843745,
          "end_to_end_ms": 1012.579074,
          "handoff_estimate_ms": 18.002156000000006,
          "handoff_plus_first_decode_ms": 61.618011,
          "output_tokens_per_s_e2e": 15.801235094455448,
          "prefill_tokens_per_s": 3005.473324718756,
          "steady_tpot_ms": 43.615854999999996,
          "steady_tpot_p95_ms": 45.577434,
          "ttft_ms": 335.388104
        },
        "prefill": {
          "gpu_active_fraction": 0.7941176470588235,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 34,
          "gross_j": {
            "core": 1.733577,
            "dram": 0.684752,
            "package-0": 8.376077,
            "psys": 13.872828,
            "uncore": 5.088305
          },
          "logits": {
            "l2": 1146.1615279083735,
            "sha256": "9992a7f0f85fa0507bed1eb81bdf57b5726e3f270c5327fd51897bcdd6e10001",
            "sum": 371186.4616530363,
            "top1": 291
          },
          "net_j": {
            "core": 0.9096894099633221,
            "dram": 0.5201493110477402,
            "package-0": 6.935271644620151,
            "psys": 10.541770082804028,
            "uncore": 5.0875926440435455
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.335388104,
          "watts": {
            "core": 5.168868481990047,
            "dram": 2.0416705059998193,
            "package-0": 24.97428173540705,
            "psys": 41.36350644088438,
            "uncore": 15.171393795171698
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 946.9425538034262,
            "sha256": "d666fa852f546efdb063aa00cdee49c7db96fa7b0654c593aeeadfd6ccfa0e85",
            "sum": 270664.63174004806,
            "top1": 42
          },
          "gpu_active_fraction": 0.9836065573770492,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 5.221788,
            "dram": 1.936213,
            "package-0": 14.337854,
            "psys": 24.578489,
            "uncore": 5.243273
          },
          "net_j": {
            "core": 3.70962070594995,
            "dram": 1.6341004232888738,
            "package-0": 11.693392835825197,
            "psys": 18.464649080714427,
            "uncore": 5.241965538409245
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.615572959,
          "token_latencies_s": [
            0.047086916,
            0.04462669,
            0.044764636,
            0.043614718,
            0.043616992,
            0.043427905,
            0.043407817,
            0.043649949,
            0.043628233,
            0.043494906,
            0.043435934,
            0.043486358,
            0.043712771,
            0.043584855
          ],
          "watts": {
            "core": 8.482809265180864,
            "dram": 3.1453834540513013,
            "package-0": 23.291884073809683,
            "psys": 39.92782437995299,
            "uncore": 8.517711707995932
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.167272,
            "dram": 2.813286,
            "package-0": 26.940544,
            "psys": 44.178109,
            "uncore": 10.910433
          },
          "net_j": {
            "core": 7.652518446906631,
            "dram": 2.3108689204163757,
            "package-0": 22.542771257002222,
            "psys": 34.01071500823536,
            "uncore": 10.908258674650007
          }
        },
        "first_decode": {
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 5,
          "gross_j": {
            "core": 1.81512,
            "dram": 0.145935,
            "package-0": 2.419123,
            "psys": 3.332755,
            "uncore": 0.30072
          },
          "logits": {
            "l2": 990.1133752307505,
            "sha256": "9ad38a6a770523e4f4bdb5ee760a99cce289de34be1c5c32497a309a9efb2956",
            "sum": 291550.6807441355,
            "top1": 78
          },
          "net_j": {
            "core": 1.6869454944095887,
            "dram": 0.120327297528879,
            "package-0": 2.1949728655507026,
            "psys": 2.8145329672945287,
            "uncore": 0.30060917678379045
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.052177269,
          "watts": {
            "core": 34.78756237701901,
            "dram": 2.7969075959111622,
            "package-0": 46.36354194773973,
            "psys": 63.87369565087817,
            "uncore": 5.76342928182002
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1000.2874800819831,
          "decode_tokens_per_s": 22.428145358317238,
          "end_to_end_ms": 1023.705705,
          "handoff_estimate_ms": 8.667740499999994,
          "handoff_plus_first_decode_ms": 52.177268999999995,
          "output_tokens_per_s_e2e": 15.629491876280985,
          "prefill_tokens_per_s": 2840.211149299021,
          "steady_tpot_ms": 43.5095285,
          "steady_tpot_p95_ms": 46.3062299,
          "ttft_ms": 354.903191
        },
        "prefill": {
          "gpu_active_fraction": 0.7714285714285715,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 2.71441,
            "dram": 0.727293,
            "package-0": 9.717443,
            "psys": 15.68569,
            "uncore": 5.350938
          },
          "logits": {
            "l2": 1146.1615279083735,
            "sha256": "9992a7f0f85fa0507bed1eb81bdf57b5726e3f270c5327fd51897bcdd6e10001",
            "sum": 371186.4616530363,
            "top1": 291
          },
          "net_j": {
            "core": 1.8425832081668683,
            "dram": 0.5531126433933821,
            "package-0": 8.192802107447061,
            "psys": 12.160809212282304,
            "uncore": 5.350184194486154
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.354903191,
          "watts": {
            "core": 7.6483110573102735,
            "dram": 2.049271515284854,
            "package-0": 27.380545586585043,
            "psys": 44.19709486353984,
            "uncore": 15.077176355960125
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 946.9425538034262,
            "sha256": "d666fa852f546efdb063aa00cdee49c7db96fa7b0654c593aeeadfd6ccfa0e85",
            "sum": 270664.63174004806,
            "top1": 42
          },
          "gpu_active_fraction": 0.9838709677419355,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 5.637742,
            "dram": 1.940058,
            "package-0": 14.803978,
            "psys": 25.159664,
            "uncore": 5.258775
          },
          "net_j": {
            "core": 4.122989744330175,
            "dram": 1.6374289794941146,
            "package-0": 12.154996284004458,
            "psys": 19.035372828658524,
            "uncore": 5.257465303380062
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.616625245,
          "token_latencies_s": [
            0.050392568,
            0.043657391,
            0.043738194,
            0.043583658,
            0.043444801,
            0.043682359,
            0.043491157,
            0.044105894,
            0.043384604,
            0.043478434,
            0.043150542,
            0.043482167,
            0.0435279,
            0.043483073
          ],
          "watts": {
            "core": 9.14289845528462,
            "dram": 3.14625133455248,
            "package-0": 24.00806343892067,
            "psys": 40.80219582965663,
            "uncore": 8.528316092540129
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.22562,
            "dram": 2.819756,
            "package-0": 26.826286,
            "psys": 44.152597,
            "uncore": 10.736239999999999
          },
          "net_j": {
            "core": 7.691431214705074,
            "dram": 2.3134559981348963,
            "package-0": 22.394525141417418,
            "psys": 33.90662446805531,
            "uncore": 10.734048870411671
          }
        },
        "first_decode": {
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6,
          "gross_j": {
            "core": 2.008113,
            "dram": 0.150024,
            "package-0": 2.60357,
            "psys": 3.597098,
            "uncore": 0.28723
          },
          "logits": {
            "l2": 990.1133752307505,
            "sha256": "9ad38a6a770523e4f4bdb5ee760a99cce289de34be1c5c32497a309a9efb2956",
            "sum": 291550.6807441355,
            "top1": 78
          },
          "net_j": {
            "core": 1.8640829965206407,
            "dram": 0.12124856248982895,
            "package-0": 2.3516919490887096,
            "psys": 3.0147706422322686,
            "uncore": 0.28710546768257283
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.058631724,
          "watts": {
            "core": 34.2495983914783,
            "dram": 2.558751299893552,
            "package-0": 44.40548260187607,
            "psys": 61.350711775079304,
            "uncore": 4.8988837510559975
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 992.6160628787688,
          "decode_tokens_per_s": 22.417209535138554,
          "end_to_end_ms": 1031.617398,
          "handoff_estimate_ms": 15.273837000000006,
          "handoff_plus_first_decode_ms": 58.631724000000006,
          "output_tokens_per_s_e2e": 15.509625982480763,
          "prefill_tokens_per_s": 2780.776944719597,
          "steady_tpot_ms": 43.357887,
          "steady_tpot_p95_ms": 44.75563305,
          "ttft_ms": 362.488621
        },
        "prefill": {
          "gpu_active_fraction": 0.7777777777777778,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 36,
          "gross_j": {
            "core": 2.569817,
            "dram": 0.734373,
            "package-0": 9.442664,
            "psys": 15.448264,
            "uncore": 5.197131
          },
          "logits": {
            "l2": 1146.1615279083735,
            "sha256": "9992a7f0f85fa0507bed1eb81bdf57b5726e3f270c5327fd51897bcdd6e10001",
            "sum": 371186.4616530363,
            "top1": 291
          },
          "net_j": {
            "core": 1.679356447554739,
            "dram": 0.5564698453574114,
            "package-0": 7.885436596367771,
            "psys": 11.848045099941755,
            "uncore": 5.196361083222282
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.362488621,
          "watts": {
            "core": 7.089372882686985,
            "dram": 2.025920146056116,
            "package-0": 26.049545980092986,
            "psys": 42.617238459466016,
            "uncore": 14.337363158221729
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 946.9425538034262,
            "sha256": "d666fa852f546efdb063aa00cdee49c7db96fa7b0654c593aeeadfd6ccfa0e85",
            "sum": 270664.63174004806,
            "top1": 42
          },
          "gpu_active_fraction": 0.9836065573770492,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 5.64769,
            "dram": 1.935359,
            "package-0": 14.780052,
            "psys": 25.107235,
            "uncore": 5.251879
          },
          "net_j": {
            "core": 4.1479917706296945,
            "dram": 1.6357375902876559,
            "package-0": 12.157396595960936,
            "psys": 19.043808725881288,
            "uncore": 5.250582319506816
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.610497053,
          "token_latencies_s": [
            0.046123105,
            0.043322607,
            0.043596273,
            0.043320962,
            0.043528307,
            0.043277703,
            0.043357916,
            0.044019302,
            0.043260678,
            0.043393456,
            0.043330948,
            0.043120893,
            0.043357858,
            0.043454986
          ],
          "watts": {
            "core": 9.25097012712361,
            "dram": 3.1701365149751184,
            "package-0": 24.20986625139368,
            "psys": 41.12589057821381,
            "uncore": 8.602627931113043
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.6549215903505683,
    "compile_target": "AUTO:GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 397440,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04699633829295635,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 13.147122000000001,
            "dram": 3.630912,
            "package-0": 38.231469,
            "psys": 62.412621,
            "uncore": 17.326982
          },
          "net_j": {
            "core": 7.6239568539536045,
            "dram": 2.9078026999115694,
            "package-0": 29.845557256631626,
            "psys": 47.54009554451513,
            "uncore": 17.324122017617896
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.388971,
            "dram": 0.162353,
            "package-0": 3.072624,
            "psys": 4.226369,
            "uncore": 0.364379
          },
          "logits": {
            "l2": 721.436499143929,
            "sha256": "33527b68f47cd3c749c20c0f2bfed74af94e75dcffa9e888e816a2d29c486e6e",
            "sum": 152204.33597268048,
            "top1": 220
          },
          "net_j": {
            "core": 2.124563591937882,
            "dram": 0.127735991323579,
            "package-0": 2.671169918930398,
            "psys": 3.5143848744053616,
            "uncore": 0.3642420856567277
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.070101923,
          "watts": {
            "core": 34.078537331993026,
            "dram": 2.315956439597242,
            "package-0": 43.830809035010354,
            "psys": 60.28891675339634,
            "uncore": 5.1978459991746595
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1398.5745163090473,
          "decode_tokens_per_s": 20.278340337242266,
          "end_to_end_ms": 1464.3481459999998,
          "handoff_estimate_ms": 23.565725499999996,
          "handoff_plus_first_decode_ms": 70.101923,
          "output_tokens_per_s_e2e": 10.926363408664432,
          "prefill_tokens_per_s": 2804.140810703604,
          "steady_tpot_ms": 46.5361975,
          "steady_tpot_p95_ms": 53.8247413,
          "ttft_ms": 724.64264
        },
        "prefill": {
          "gpu_active_fraction": 0.8194444444444444,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 72,
          "gross_j": {
            "core": 4.563831,
            "dram": 1.478451,
            "package-0": 19.339733,
            "psys": 31.46043,
            "uncore": 11.421479
          },
          "logits": {
            "l2": 834.8475507956192,
            "sha256": "765152f2d994661cc75a99376a54eb9c804048cab9a622294153aa34f8ebd025",
            "sum": 210672.0778277011,
            "top1": 220
          },
          "net_j": {
            "core": 1.8306551609621668,
            "dram": 1.1206154445305068,
            "package-0": 15.1899075359389,
            "psys": 24.100659620962176,
            "uncore": 11.420063718265679
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.72464264,
          "watts": {
            "core": 6.298043681227481,
            "dram": 2.0402484181720246,
            "package-0": 26.688648904237816,
            "psys": 43.41509630181299,
            "uncore": 15.761533160676274
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 677.6366610405684,
            "sha256": "ad393a0a123b7a19df0f9f33da818df6a7780c377cdd4707a8b74388cd4bd9a9",
            "sum": 116901.59847021359,
            "top1": 42
          },
          "gpu_active_fraction": 0.9848484848484849,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 66,
          "gross_j": {
            "core": 6.19432,
            "dram": 1.990108,
            "package-0": 15.819112,
            "psys": 26.725822,
            "uncore": 5.541124
          },
          "net_j": {
            "core": 3.668738101053556,
            "dram": 1.6594512640574837,
            "package-0": 11.984479801762328,
            "psys": 19.925051049147598,
            "uncore": 5.53981621369549
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.669603583,
          "token_latencies_s": [
            0.052774091,
            0.055775949,
            0.047196928,
            0.046725191,
            0.046533828,
            0.046494774,
            0.046538567,
            0.046446849,
            0.046454437,
            0.046180024,
            0.046253406,
            0.046389088,
            0.04854073,
            0.047275331
          ],
          "watts": {
            "core": 9.25072708280296,
            "dram": 2.972068923352819,
            "package-0": 23.62459282121255,
            "psys": 39.91290172053933,
            "uncore": 8.275230510527301
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.10867,
            "dram": 3.71929,
            "package-0": 38.512414,
            "psys": 63.272422000000006,
            "uncore": 17.464005
          },
          "net_j": {
            "core": 7.454434210195799,
            "dram": 2.97902053921319,
            "package-0": 29.92749555999388,
            "psys": 48.04695553438617,
            "uncore": 17.461077147162815
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 1.861568,
            "dram": 0.153137,
            "package-0": 2.503961,
            "psys": 3.4887,
            "uncore": 0.320983
          },
          "logits": {
            "l2": 721.436499143929,
            "sha256": "33527b68f47cd3c749c20c0f2bfed74af94e75dcffa9e888e816a2d29c486e6e",
            "sum": 152204.33597268048,
            "top1": 220
          },
          "net_j": {
            "core": 1.631492235427416,
            "dram": 0.12301479277132733,
            "package-0": 2.154633213309647,
            "psys": 2.8691625473230142,
            "uncore": 0.32086386312543136
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.060999628,
          "watts": {
            "core": 30.51769430462756,
            "dram": 2.5104579326287038,
            "package-0": 41.04879131394047,
            "psys": 57.19215205705845,
            "uncore": 5.262048483312062
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1366.154208948247,
          "decode_tokens_per_s": 20.844968010130874,
          "end_to_end_ms": 1499.0987010000001,
          "handoff_estimate_ms": 14.474234500000001,
          "handoff_plus_first_decode_ms": 60.999628,
          "output_tokens_per_s_e2e": 10.67307975740818,
          "prefill_tokens_per_s": 2606.797320742437,
          "steady_tpot_ms": 46.5253935,
          "steady_tpot_p95_ms": 49.796110549999995,
          "ttft_ms": 779.500571
        },
        "prefill": {
          "gpu_active_fraction": 0.7922077922077922,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 77,
          "gross_j": {
            "core": 5.351366,
            "dram": 1.581722,
            "package-0": 20.582589,
            "psys": 33.612402,
            "uncore": 11.631989
          },
          "logits": {
            "l2": 834.8475507956192,
            "sha256": "765152f2d994661cc75a99376a54eb9c804048cab9a622294153aa34f8ebd025",
            "sum": 210672.0778277011,
            "top1": 220
          },
          "net_j": {
            "core": 2.411279384101444,
            "dram": 1.1967970667769536,
            "package-0": 16.118607527993372,
            "psys": 25.69547178335022,
            "uncore": 11.630466576588608
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.779500571,
          "watts": {
            "core": 6.865121334208746,
            "dram": 2.0291479683855163,
            "package-0": 26.40484146611356,
            "psys": 43.12043281364062,
            "uncore": 14.922361102414126
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 677.6366610405684,
            "sha256": "ad393a0a123b7a19df0f9f33da818df6a7780c377cdd4707a8b74388cd4bd9a9",
            "sum": 116901.59847021359,
            "top1": 42
          },
          "gpu_active_fraction": 0.9848484848484849,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 66,
          "gross_j": {
            "core": 5.895736,
            "dram": 1.984431,
            "package-0": 15.425864,
            "psys": 26.17132,
            "uncore": 5.511033
          },
          "net_j": {
            "core": 3.411662590666939,
            "dram": 1.659208679664909,
            "package-0": 11.654254818690859,
            "psys": 19.482321203712935,
            "uncore": 5.509746707448775
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.658598502,
          "token_latencies_s": [
            0.052068207,
            0.048572674,
            0.046855646,
            0.046603811,
            0.046443309,
            0.046609289,
            0.046416766,
            0.046500834,
            0.046319965,
            0.046558583,
            0.046549953,
            0.046453821,
            0.046215314,
            0.046396657
          ],
          "watts": {
            "core": 8.951942620725852,
            "dram": 3.013111924751994,
            "package-0": 23.42225795102097,
            "psys": 39.73789785510323,
            "uncore": 8.367818911316018
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.061609,
            "dram": 3.7170330000000003,
            "package-0": 39.644611999999995,
            "psys": 64.618244,
            "uncore": 17.620744
          },
          "net_j": {
            "core": 8.384942202594605,
            "dram": 2.9738268046094114,
            "package-0": 31.025636195224855,
            "psys": 49.33237634725531,
            "uncore": 17.61780453203184
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.066157,
            "dram": 0.155701,
            "package-0": 2.720513,
            "psys": 3.73046,
            "uncore": 0.306029
          },
          "logits": {
            "l2": 721.436499143929,
            "sha256": "33527b68f47cd3c749c20c0f2bfed74af94e75dcffa9e888e816a2d29c486e6e",
            "sum": 152204.33597268048,
            "top1": 220
          },
          "net_j": {
            "core": 1.8192842632141077,
            "dram": 0.12337968293129936,
            "package-0": 2.345682096380861,
            "psys": 3.06569243194319,
            "uncore": 0.3059011653872083
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.065452983,
          "watts": {
            "core": 31.56704103157529,
            "dram": 2.378822062242755,
            "package-0": 41.56438523206803,
            "psys": 56.99449939508486,
            "uncore": 4.675554664941703
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1360.7559327169686,
          "decode_tokens_per_s": 20.416328324438073,
          "end_to_end_ms": 1505.045799,
          "handoff_estimate_ms": 18.99911950000001,
          "handoff_plus_first_decode_ms": 65.452983,
          "output_tokens_per_s_e2e": 10.630905724351317,
          "prefill_tokens_per_s": 2637.797167485111,
          "steady_tpot_ms": 46.4538635,
          "steady_tpot_p95_ms": 54.70955785,
          "ttft_ms": 770.339746
        },
        "prefill": {
          "gpu_active_fraction": 0.7792207792207793,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 77,
          "gross_j": {
            "core": 5.527451,
            "dram": 1.574337,
            "package-0": 20.859626,
            "psys": 33.895238,
            "uncore": 11.754304
          },
          "logits": {
            "l2": 834.8475507956192,
            "sha256": "765152f2d994661cc75a99376a54eb9c804048cab9a622294153aa34f8ebd025",
            "sum": 210672.0778277011,
            "top1": 220
          },
          "net_j": {
            "core": 2.621916788519233,
            "dram": 1.1939357709057004,
            "package-0": 16.448106007291372,
            "psys": 26.071348911157088,
            "uncore": 11.752799468370815
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.770339746,
          "watts": {
            "core": 7.1753418263842255,
            "dram": 2.043691771292819,
            "package-0": 27.078475579526955,
            "psys": 44.00037538761501,
            "uncore": 15.258597341023085
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 677.6366610405684,
            "sha256": "ad393a0a123b7a19df0f9f33da818df6a7780c377cdd4707a8b74388cd4bd9a9",
            "sum": 116901.59847021359,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 67,
          "gross_j": {
            "core": 6.468001,
            "dram": 1.986995,
            "package-0": 16.064473,
            "psys": 26.992546,
            "uncore": 5.560411
          },
          "net_j": {
            "core": 3.9437411508612636,
            "dram": 1.6565113507724116,
            "package-0": 12.231848091552623,
            "psys": 20.195335004155034,
            "uncore": 5.5591038982738175
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.66925307,
          "token_latencies_s": [
            0.057846764,
            0.053020293,
            0.046569169,
            0.046005558,
            0.046458924,
            0.046382506,
            0.048290515,
            0.046751342,
            0.046448803,
            0.047179591,
            0.045928717,
            0.046342415,
            0.046035256,
            0.045965908
          ],
          "watts": {
            "core": 9.664507030203088,
            "dram": 2.968974053417491,
            "package-0": 24.003585071339305,
            "psys": 40.33234543100415,
            "uncore": 8.308383254782829
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 4.149856457486749,
    "compile_target": "NPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 397509,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 2048,
      "NPUW_LLM_MIN_RESPONSE_LEN": 2048,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04626445472240448,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 14.136012000000001,
            "dram": 4.244863,
            "package-0": 37.968593,
            "psys": 65.43330399999999,
            "uncore": 12.25943
          },
          "net_j": {
            "core": 7.400698304893958,
            "dram": 3.3630555864682017,
            "package-0": 27.74225634021526,
            "psys": 47.29676059918477,
            "uncore": 12.257339588971764
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.04,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 25,
          "gross_j": {
            "core": 3.720999,
            "dram": 0.290161,
            "package-0": 4.528492,
            "psys": 7.148724,
            "uncore": 0.011536
          },
          "logits": {
            "l2": 1042.713244693195,
            "sha256": "38e7cb73dd1e63942c1fc408d3f6c090b083c85b5b9625f370be95e69b4098fc",
            "sum": 237734.52709483728,
            "top1": 13
          },
          "net_j": {
            "core": 2.7860254105715496,
            "dram": 0.16775146127079268,
            "package-0": 3.1089063464839537,
            "psys": 4.631070056783312,
            "uncore": 0.011051855868367108
          },
          "npu_busy_delta_us": 68627,
          "npu_utilization": 0.2768467046179593,
          "seconds": 0.247888087,
          "watts": {
            "core": 15.010802031805586,
            "dram": 1.1705322490951329,
            "package-0": 18.268292174928114,
            "psys": 28.838513728172824,
            "uncore": 0.04653712947488275
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1146.8742767303595,
          "decode_tokens_per_s": 15.57157535236046,
          "end_to_end_ms": 1785.7231969999998,
          "handoff_estimate_ms": 198.79146350000002,
          "handoff_plus_first_decode_ms": 247.888087,
          "output_tokens_per_s_e2e": 8.959955286955934,
          "prefill_tokens_per_s": 2470.728418948724,
          "steady_tpot_ms": 49.0966235,
          "steady_tpot_p95_ms": 59.35882599999999,
          "ttft_ms": 822.429525
        },
        "prefill": {
          "gpu_active_fraction": 0.9878048780487805,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 82,
          "gross_j": {
            "core": 6.038193,
            "dram": 1.849239,
            "package-0": 22.580265,
            "psys": 38.103784,
            "uncore": 12.247894
          },
          "logits": {
            "l2": 3172.3464720715606,
            "sha256": "f004143b6702c8233fb4ece6ff62551d0fe6c4f0eacaff333650b159de35221c",
            "sum": 1184720.837936934,
            "top1": 283
          },
          "net_j": {
            "core": 2.9361888079180076,
            "dram": 1.443115333425692,
            "package-0": 17.870441430632518,
            "psys": 29.750849570222105,
            "uncore": 12.246287733103397
          },
          "npu_busy_delta_us": 9040,
          "npu_utilization": 0.010991823281149834,
          "seconds": 0.822429525,
          "watts": {
            "core": 7.341897167419908,
            "dram": 2.2485075544922832,
            "package-0": 27.455562225833273,
            "psys": 46.33075885742307,
            "uncore": 14.892332567948603
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 911.2120236742918,
            "sha256": "d6f9dc82df9060085096789dc10c8db0740cc84a5a4516f697733fbdfebb4735",
            "sum": 196145.01117703877,
            "top1": 271
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 71,
          "gross_j": {
            "core": 4.37682,
            "dram": 2.105463,
            "package-0": 10.859836,
            "psys": 20.180796,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.678484086404401,
            "dram": 1.752188791771717,
            "package-0": 6.762908563098788,
            "psys": 12.914840972179356,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 695949,
          "npu_utilization": 0.9728034203143663,
          "seconds": 0.715405585,
          "token_latencies_s": [
            0.069827999,
            0.053721579,
            0.051255127,
            0.0500679,
            0.049292558,
            0.048827563,
            0.048992746,
            0.049012211,
            0.049140611,
            0.049001625,
            0.049052636,
            0.049228054,
            0.048985064,
            0.048972996
          ],
          "watts": {
            "core": 6.117956152103566,
            "dram": 2.9430340552904686,
            "package-0": 15.17997095312025,
            "psys": 28.208887969472592,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.448879999999999,
            "dram": 4.13683,
            "package-0": 37.079495,
            "psys": 62.856467,
            "uncore": 12.373565000000001
          },
          "net_j": {
            "core": 7.373549958693754,
            "dram": 3.341429620360823,
            "package-0": 27.855222288419426,
            "psys": 46.497097272575225,
            "uncore": 12.371799312282544
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 9,
          "gross_j": {
            "core": 2.361994,
            "dram": 0.184082,
            "package-0": 2.832817,
            "psys": 4.046986,
            "uncore": 0.006104
          },
          "logits": {
            "l2": 1135.1989390721888,
            "sha256": "d7d9a69882fb9e20540d9d9278969b3150f2fc54597af1b7836d34e8fe5c7737",
            "sum": 279393.15176730603,
            "top1": 819
          },
          "net_j": {
            "core": 2.027999367858062,
            "dram": 0.14035442381945956,
            "package-0": 2.3257075013991866,
            "psys": 3.1476205931767027,
            "uncore": 0.005931052290057462
          },
          "npu_busy_delta_us": 63352,
          "npu_utilization": 0.7154256888436924,
          "seconds": 0.088551475,
          "watts": {
            "core": 26.673683301153368,
            "dram": 2.078813481085436,
            "package-0": 31.990624662096252,
            "psys": 45.70207328562286,
            "uncore": 0.06893165811184963
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1271.4631090175278,
          "decode_tokens_per_s": 18.862249719637806,
          "end_to_end_ms": 1610.742762,
          "handoff_estimate_ms": 39.3654405,
          "handoff_plus_first_decode_ms": 88.55147500000001,
          "output_tokens_per_s_e2e": 9.933305539199436,
          "prefill_tokens_per_s": 2491.7119248165363,
          "steady_tpot_ms": 49.186034500000005,
          "steady_tpot_p95_ms": 56.62088194999999,
          "ttft_ms": 815.5035819999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9876543209876543,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 81,
          "gross_j": {
            "core": 5.98143,
            "dram": 1.856501,
            "package-0": 22.647769,
            "psys": 37.993555,
            "uncore": 12.367461
          },
          "logits": {
            "l2": 3189.622647004578,
            "sha256": "382d46debf88546ab06200f51b873649d7eb5555857d5baa3b57d12ba7070aec",
            "sum": 1191507.7713742703,
            "top1": 283
          },
          "net_j": {
            "core": 2.905548780850091,
            "dram": 1.453797431172845,
            "package-0": 17.977608366592285,
            "psys": 29.710963312043518,
            "uncore": 12.365868259992487
          },
          "npu_busy_delta_us": 8549,
          "npu_utilization": 0.010483093132508154,
          "seconds": 0.815503582,
          "watts": {
            "core": 7.334645894909141,
            "dram": 2.27650870085326,
            "package-0": 27.77151382273144,
            "psys": 46.589071879760304,
            "uncore": 15.165428175887524
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1211.5266628791376,
            "sha256": "f06f3f53b14c831c609644c43edf15f5e077fe256d545be2836d7ea42cae549c",
            "sum": 348738.0987180695,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 70,
          "gross_j": {
            "core": 5.105456,
            "dram": 2.096247,
            "package-0": 11.598909,
            "psys": 20.815926,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.440001809985601,
            "dram": 1.7472777653685183,
            "package-0": 7.551906420427954,
            "psys": 13.638513367355007,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 691596,
          "npu_utilization": 0.9786444494601756,
          "seconds": 0.706687705,
          "token_latencies_s": [
            0.060445064,
            0.054561707,
            0.050538984,
            0.049638074,
            0.049234393,
            0.049137676,
            0.049290505,
            0.049115283,
            0.049128557,
            0.049092839,
            0.049116978,
            0.049324404,
            0.049071654,
            0.048969642
          ],
          "watts": {
            "core": 7.2244868049600495,
            "dram": 2.966298953793175,
            "package-0": 16.413061834718068,
            "psys": 29.455622126608247,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.824611,
            "dram": 4.131642,
            "package-0": 37.298122,
            "psys": 62.981955000000006,
            "uncore": 12.24124
          },
          "net_j": {
            "core": 7.7169512128683255,
            "dram": 3.332008913342711,
            "package-0": 28.024762507921537,
            "psys": 46.5355292177509,
            "uncore": 12.239458946692075
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.1111111111111111,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 9,
          "gross_j": {
            "core": 2.377924,
            "dram": 0.182983,
            "package-0": 2.842461,
            "psys": 4.053334,
            "uncore": 0.002746
          },
          "logits": {
            "l2": 1352.6577453890102,
            "sha256": "e62f0e0966003b386b36bde47a4495306738b10dbb4a67af95c88e4a6757f69a",
            "sum": 376909.0747744143,
            "top1": 13
          },
          "net_j": {
            "core": 2.0424688568565017,
            "dram": 0.13906420935091054,
            "package-0": 2.333133983719479,
            "psys": 3.1500357962252536,
            "uncore": 0.002572296014330975
          },
          "npu_busy_delta_us": 63344,
          "npu_utilization": 0.7122209052351992,
          "seconds": 0.088938698,
          "watts": {
            "core": 26.73666304402163,
            "dram": 2.0574058774730433,
            "package-0": 31.959777508773517,
            "psys": 45.574469731949534,
            "uncore": 0.030875199005049526
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1264.732858713221,
          "decode_tokens_per_s": 18.836398888604243,
          "end_to_end_ms": 1619.3142970000001,
          "handoff_estimate_ms": 39.74441449999999,
          "handoff_plus_first_decode_ms": 88.938698,
          "output_tokens_per_s_e2e": 9.88072545869704,
          "prefill_tokens_per_s": 2469.0645741444846,
          "steady_tpot_ms": 49.194283500000004,
          "steady_tpot_p95_ms": 56.315508099999995,
          "ttft_ms": 822.983741
        },
        "prefill": {
          "gpu_active_fraction": 0.9878048780487805,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 82,
          "gross_j": {
            "core": 6.502852,
            "dram": 1.85052,
            "package-0": 23.020082,
            "psys": 38.300378,
            "uncore": 12.238494
          },
          "logits": {
            "l2": 3199.9982517626586,
            "sha256": "4900c9dd8e9065590b01c6d5567d6284f9f760d8766220320e786ff3db59567a",
            "sum": 1195583.9069854915,
            "top1": 283
          },
          "net_j": {
            "core": 3.3987574399252987,
            "dram": 1.4441226561960299,
            "package-0": 18.307084590807982,
            "psys": 29.941814723290896,
            "uncore": 12.236886650677745
          },
          "npu_busy_delta_us": 8781,
          "npu_utilization": 0.010669712611005277,
          "seconds": 0.822983741,
          "watts": {
            "core": 7.901555858319198,
            "dram": 2.2485498896387064,
            "package-0": 27.971490629971022,
            "psys": 46.538438236290744,
            "uncore": 14.870881878090467
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1248.264823861332,
            "sha256": "6193653dd623285c9088f54e32bb5cecc2517a2fd6a7aaa09c41215c455b3de1",
            "sum": 317309.1419824492,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 70,
          "gross_j": {
            "core": 4.943835,
            "dram": 2.098139,
            "package-0": 11.435579,
            "psys": 20.628243,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.275724916086525,
            "dram": 1.7488220477957706,
            "package-0": 7.384543933394076,
            "psys": 13.443678698234748,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 691105,
          "npu_utilization": 0.9769761867968799,
          "seconds": 0.707391858,
          "token_latencies_s": [
            0.063301686,
            0.05255372,
            0.050674762,
            0.049584771,
            0.049100925,
            0.049055602,
            0.049253311,
            0.048966893,
            0.049053842,
            0.048887522,
            0.049193885,
            0.049140436,
            0.049194682,
            0.049404227
          ],
          "watts": {
            "core": 6.98882089762475,
            "dram": 2.9660208500731713,
            "package-0": 16.16583350610179,
            "psys": 29.16098449072056,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.34754111431539,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 397355,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06940924748778343,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 13.624232,
            "dram": 3.6068029999999998,
            "package-0": 39.672506,
            "psys": 67.034192,
            "uncore": 17.298112
          },
          "net_j": {
            "core": 8.14529282271186,
            "dram": 2.889483896016317,
            "package-0": 31.3537432664774,
            "psys": 52.28075620212233,
            "uncore": 17.295274918524917
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.009577,
            "dram": 0.15741,
            "package-0": 2.757867,
            "psys": 3.983877,
            "uncore": 0.360167
          },
          "logits": {
            "l2": 721.436499143929,
            "sha256": "33527b68f47cd3c749c20c0f2bfed74af94e75dcffa9e888e816a2d29c486e6e",
            "sum": 152204.33597268048,
            "top1": 220
          },
          "net_j": {
            "core": 1.7566192162294858,
            "dram": 0.12429201038037897,
            "package-0": 2.3737970702757267,
            "psys": 3.3027238968770525,
            "uncore": 0.3600360144535118
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.067066302,
          "watts": {
            "core": 29.96403469509919,
            "dram": 2.3470803563912024,
            "package-0": 41.12150092903587,
            "psys": 59.40206752416438,
            "uncore": 5.370312500605745
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1409.8638025856274,
          "decode_tokens_per_s": 20.57018616979797,
          "end_to_end_ms": 1452.622584,
          "handoff_estimate_ms": 20.334507499999994,
          "handoff_plus_first_decode_ms": 67.066302,
          "output_tokens_per_s_e2e": 11.014560957700214,
          "prefill_tokens_per_s": 2808.9115821912337,
          "steady_tpot_ms": 46.7317945,
          "steady_tpot_p95_ms": 49.735279850000005,
          "ttft_ms": 723.411877
        },
        "prefill": {
          "gpu_active_fraction": 0.8194444444444444,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 72,
          "gross_j": {
            "core": 5.103136,
            "dram": 1.472653,
            "package-0": 20.405221,
            "psys": 34.211216,
            "uncore": 11.43094
          },
          "logits": {
            "l2": 834.8475507956192,
            "sha256": "765152f2d994661cc75a99376a54eb9c804048cab9a622294153aa34f8ebd025",
            "sum": 210672.0778277011,
            "top1": 220
          },
          "net_j": {
            "core": 2.3746022997344336,
            "dram": 1.115425207222689,
            "package-0": 16.262443770966343,
            "psys": 26.863945759059234,
            "uncore": 11.429527122038651
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.723411877,
          "watts": {
            "core": 7.054260736169804,
            "dram": 2.035704757996391,
            "package-0": 28.20692008074399,
            "psys": 47.29147680277857,
            "uncore": 15.801427047900127
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 677.6366610405684,
            "sha256": "ad393a0a123b7a19df0f9f33da818df6a7780c377cdd4707a8b74388cd4bd9a9",
            "sum": 116901.59847021359,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 66,
          "gross_j": {
            "core": 6.511519,
            "dram": 1.97674,
            "package-0": 16.509418,
            "psys": 28.839099,
            "uncore": 5.507005
          },
          "net_j": {
            "core": 4.014071306747941,
            "dram": 1.6497666784132492,
            "package-0": 12.717502425235333,
            "psys": 22.114086546186044,
            "uncore": 5.505711782032756
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.662144405,
          "token_latencies_s": [
            0.051925592,
            0.047355167,
            0.047225469,
            0.046555757,
            0.048555881,
            0.047104171,
            0.046852493,
            0.046531202,
            0.046263863,
            0.046176471,
            0.046450101,
            0.046191371,
            0.046611096,
            0.04831796
          ],
          "watts": {
            "core": 9.833986288836797,
            "dram": 2.9853608745663265,
            "package-0": 24.933259082661884,
            "psys": 43.55409300785377,
            "uncore": 8.316924463025554
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.77218,
            "dram": 3.732962,
            "package-0": 40.64516,
            "psys": 65.94660900000001,
            "uncore": 17.642106
          },
          "net_j": {
            "core": 9.03250984034242,
            "dram": 2.981507215987359,
            "package-0": 31.930525162512343,
            "psys": 50.49108878679413,
            "uncore": 17.639133907893093
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.026118,
            "dram": 0.155761,
            "package-0": 2.665338,
            "psys": 3.70098,
            "uncore": 0.312316
          },
          "logits": {
            "l2": 721.436499143929,
            "sha256": "33527b68f47cd3c749c20c0f2bfed74af94e75dcffa9e888e816a2d29c486e6e",
            "sum": 152204.33597268048,
            "top1": 220
          },
          "net_j": {
            "core": 1.785999250134327,
            "dram": 0.12432393509931253,
            "package-0": 2.3007617849029613,
            "psys": 3.0543992577470283,
            "uncore": 0.3121916627085974
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.063662309,
          "watts": {
            "core": 31.82602126479578,
            "dram": 2.446675316159205,
            "package-0": 41.86681321910583,
            "psys": 58.1345549373649,
            "uncore": 4.905822690157216
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1345.8191512328403,
          "decode_tokens_per_s": 20.532744196156006,
          "end_to_end_ms": 1521.749782,
          "handoff_estimate_ms": 16.6687645,
          "handoff_plus_first_decode_ms": 63.662309,
          "output_tokens_per_s_e2e": 10.514212119006565,
          "prefill_tokens_per_s": 2568.2204376300456,
          "steady_tpot_ms": 46.9935445,
          "steady_tpot_p95_ms": 50.317366,
          "ttft_ms": 791.209341
        },
        "prefill": {
          "gpu_active_fraction": 0.759493670886076,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 79,
          "gross_j": {
            "core": 6.144089,
            "dram": 1.587582,
            "package-0": 21.795537,
            "psys": 35.291535,
            "uncore": 11.809785
          },
          "logits": {
            "l2": 834.8475507956192,
            "sha256": "765152f2d994661cc75a99376a54eb9c804048cab9a622294153aa34f8ebd025",
            "sum": 210672.0778277011,
            "top1": 220
          },
          "net_j": {
            "core": 3.1598397506841636,
            "dram": 1.1968751627244754,
            "package-0": 17.264502681229025,
            "psys": 27.25568567224055,
            "uncore": 11.808239708477627
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.791209341,
          "watts": {
            "core": 7.765440423434031,
            "dram": 2.006525855715346,
            "package-0": 27.547117899863117,
            "psys": 44.604547963748075,
            "uncore": 14.926245669791708
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 677.6366610405684,
            "sha256": "ad393a0a123b7a19df0f9f33da818df6a7780c377cdd4707a8b74388cd4bd9a9",
            "sum": 116901.59847021359,
            "top1": 42
          },
          "gpu_active_fraction": 0.9848484848484849,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 66,
          "gross_j": {
            "core": 6.601973,
            "dram": 1.989619,
            "package-0": 16.184285,
            "psys": 26.954094,
            "uncore": 5.520005
          },
          "net_j": {
            "core": 4.08667083952393,
            "dram": 1.660308118163571,
            "package-0": 12.365260696380357,
            "psys": 20.181003856806555,
            "uncore": 5.5187025367068685
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.666878132,
          "token_latencies_s": [
            0.053377475,
            0.048602041,
            0.047395822,
            0.047230073,
            0.046868216,
            0.046658153,
            0.046766056,
            0.046636233,
            0.04684158,
            0.048669615,
            0.047226088,
            0.047110818,
            0.046876271,
            0.046578755
          ],
          "watts": {
            "core": 9.899819297118592,
            "dram": 2.983482145430434,
            "package-0": 24.268729507537667,
            "psys": 40.41832038960906,
            "uncore": 8.277381931006248
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.257594,
            "dram": 3.71868,
            "package-0": 38.737999,
            "psys": 63.272544,
            "uncore": 17.533768
          },
          "net_j": {
            "core": 7.538064301499767,
            "dram": 2.969862065399941,
            "package-0": 30.053943753248852,
            "psys": 47.87125709565368,
            "uncore": 17.53080633694284
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.093073,
            "dram": 0.156494,
            "package-0": 2.715142,
            "psys": 3.769705,
            "uncore": 0.320373
          },
          "logits": {
            "l2": 721.436499143929,
            "sha256": "33527b68f47cd3c749c20c0f2bfed74af94e75dcffa9e888e816a2d29c486e6e",
            "sum": 152204.33597268048,
            "top1": 220
          },
          "net_j": {
            "core": 1.8379431897010392,
            "dram": 0.12309164217098682,
            "package-0": 2.3277742469642435,
            "psys": 3.0827031635909528,
            "uncore": 0.3202408897437773
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.067642168,
          "watts": {
            "core": 30.943316305296424,
            "dram": 2.313556833364655,
            "package-0": 40.13978381059578,
            "psys": 55.73010315104034,
            "uncore": 4.736291125382025
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1350.5582503841788,
          "decode_tokens_per_s": 20.358070335927813,
          "end_to_end_ms": 1516.409973,
          "handoff_estimate_ms": 21.047647,
          "handoff_plus_first_decode_ms": 67.642168,
          "output_tokens_per_s_e2e": 10.551236331126397,
          "prefill_tokens_per_s": 2606.460049405989,
          "steady_tpot_ms": 46.594521,
          "steady_tpot_p95_ms": 53.8099788,
          "ttft_ms": 779.6014369999999
        },
        "prefill": {
          "gpu_active_fraction": 0.7922077922077922,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 77,
          "gross_j": {
            "core": 5.331956,
            "dram": 1.576229,
            "package-0": 20.611947,
            "psys": 33.352271,
            "uncore": 11.681244
          },
          "logits": {
            "l2": 834.8475507956192,
            "sha256": "765152f2d994661cc75a99376a54eb9c804048cab9a622294153aa34f8ebd025",
            "sum": 210672.0778277011,
            "top1": 220
          },
          "net_j": {
            "core": 2.3914889420753167,
            "dram": 1.1912542581690335,
            "package-0": 16.147387896658767,
            "psys": 25.434316346533954,
            "uncore": 11.6797213795897
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.779601437,
          "watts": {
            "core": 6.839335777160708,
            "dram": 2.0218395261885598,
            "package-0": 26.439082872034266,
            "psys": 42.781182046487174,
            "uncore": 14.983610144371758
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 677.6366610405684,
            "sha256": "ad393a0a123b7a19df0f9f33da818df6a7780c377cdd4707a8b74388cd4bd9a9",
            "sum": 116901.59847021359,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 66,
          "gross_j": {
            "core": 5.832565,
            "dram": 1.985957,
            "package-0": 15.41091,
            "psys": 26.150568,
            "uncore": 5.532151
          },
          "net_j": {
            "core": 3.3086321697234107,
            "dram": 1.6555161650599208,
            "package-0": 11.578781609625842,
            "psys": 19.354237585528775,
            "uncore": 5.530844067609363
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.669166368,
          "token_latencies_s": [
            0.057696464,
            0.051717256,
            0.047029774,
            0.046733732,
            0.048788985,
            0.046665355,
            0.046676283,
            0.046523687,
            0.046197304,
            0.046466676,
            0.046125517,
            0.046097711,
            0.046198812,
            0.046221872
          ],
          "watts": {
            "core": 8.716165783155438,
            "dram": 2.967807551260556,
            "package-0": 23.03001276955987,
            "psys": 39.079322049849345,
            "uncore": 8.267228098349378
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.671638049185276,
    "compile_target": "AUTO:GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 397842,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04640396311879158,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 13.511133999999998,
            "dram": 5.485217,
            "package-0": 59.912627,
            "psys": 99.782277,
            "uncore": 34.612826
          },
          "net_j": {
            "core": 4.442110618724019,
            "dram": 4.297873334490673,
            "package-0": 46.14298085274251,
            "psys": 75.36162733122013,
            "uncore": 34.608129916068506
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.309564,
            "dram": 0.167969,
            "package-0": 3.033867,
            "psys": 4.224049,
            "uncore": 0.359802
          },
          "logits": {
            "l2": 979.727719530536,
            "sha256": "8eca8a515a86885daba895c0e4490a5dfd1090aa41c110d1f8b0c9ea55c06524",
            "sum": 284299.187461914,
            "top1": 78
          },
          "net_j": {
            "core": 2.0220074682353184,
            "dram": 0.13032123881200441,
            "package-0": 2.5972652267696854,
            "psys": 3.4497299772160224,
            "uncore": 0.3596530986738656
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.076239414,
          "watts": {
            "core": 30.29356967512893,
            "dram": 2.2031780044899087,
            "package-0": 39.79394437633007,
            "psys": 55.405055972754454,
            "uncore": 4.719369957381886
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1703.5038278928787,
          "decode_tokens_per_s": 21.21861532151427,
          "end_to_end_ms": 2404.456,
          "handoff_estimate_ms": 31.544564000000005,
          "handoff_plus_first_decode_ms": 76.23941400000001,
          "output_tokens_per_s_e2e": 6.654311827706557,
          "prefill_tokens_per_s": 2403.4927429296517,
          "steady_tpot_ms": 44.69485,
          "steady_tpot_p95_ms": 46.5722248,
          "ttft_ms": 1697.529569
        },
        "prefill": {
          "gpu_active_fraction": 0.844311377245509,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 167,
          "gross_j": {
            "core": 6.689741,
            "dram": 3.32714,
            "package-0": 42.845044,
            "psys": 71.014589,
            "uncore": 28.685656
          },
          "logits": {
            "l2": 1150.3575819756547,
            "sha256": "2c23050136374b58161b79874f63495a8c895d72d93c7555b2aaf74e81ff149c",
            "sum": 369774.01646265644,
            "top1": 220
          },
          "net_j": {
            "core": 0.2870722251634721,
            "dram": 2.4888834543335676,
            "package-0": 33.12376754926166,
            "psys": 53.773776563863144,
            "uncore": 28.682340596710354
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.697529569,
          "watts": {
            "core": 3.940868614112488,
            "dram": 1.959989422723275,
            "package-0": 25.23964517757393,
            "psys": 41.83408071167448,
            "uncore": 16.89847206425893
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 811.0208018948907,
            "sha256": "b95537717dfcf6809ff402f9c640cbc97a728f4a0d1eb86875f847d4d31b12aa",
            "sum": 202420.8223354963,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 4.511829,
            "dram": 1.990108,
            "package-0": 14.033716,
            "psys": 24.543639,
            "uncore": 5.567368
          },
          "net_j": {
            "core": 2.1330309253252286,
            "dram": 1.6786686413451013,
            "package-0": 10.42194807671117,
            "psys": 18.138120790140967,
            "uncore": 5.56613622068429
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.630687017,
          "token_latencies_s": [
            0.049076706,
            0.044911002,
            0.044890823,
            0.044562977,
            0.044629914,
            0.044604457,
            0.044795915,
            0.044648014,
            0.044590616,
            0.044683985,
            0.045223658,
            0.044705715,
            0.044758398,
            0.044575731
          ],
          "watts": {
            "core": 7.153832056764853,
            "dram": 3.1554605475571416,
            "package-0": 22.251474379089686,
            "psys": 38.915719427279726,
            "uncore": 8.82746568413981
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 19.837414,
            "dram": 5.683273,
            "package-0": 67.079846,
            "psys": 108.62521100000001,
            "uncore": 34.791049
          },
          "net_j": {
            "core": 10.272295710268732,
            "dram": 4.430979105311428,
            "package-0": 52.55697086019312,
            "psys": 82.86869974474936,
            "uncore": 34.78609603026833
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 1.954402,
            "dram": 0.163085,
            "package-0": 2.62481,
            "psys": 3.706778,
            "uncore": 0.321899
          },
          "logits": {
            "l2": 979.727719530536,
            "sha256": "8eca8a515a86885daba895c0e4490a5dfd1090aa41c110d1f8b0c9ea55c06524",
            "sum": 284299.187461914,
            "top1": 78
          },
          "net_j": {
            "core": 1.6805868119871508,
            "dram": 0.12723629663481364,
            "package-0": 2.2090719331574333,
            "psys": 2.969461036603652,
            "uncore": 0.3217572141592797
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.072596193,
          "watts": {
            "core": 26.921549453702067,
            "dram": 2.2464676625673747,
            "package-0": 36.15630367834853,
            "psys": 51.060225706325944,
            "uncore": 4.434103038984427
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1615.1515932468096,
          "decode_tokens_per_s": 21.282415866569966,
          "end_to_end_ms": 2535.984868,
          "handoff_estimate_ms": 27.9302605,
          "handoff_plus_first_decode_ms": 72.596193,
          "output_tokens_per_s_e2e": 6.30918591112035,
          "prefill_tokens_per_s": 2228.0743578204356,
          "steady_tpot_ms": 44.665932500000004,
          "steady_tpot_p95_ms": 47.106267450000004,
          "ttft_ms": 1831.177665
        },
        "prefill": {
          "gpu_active_fraction": 0.8066298342541437,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 181,
          "gross_j": {
            "core": 12.734587,
            "dram": 3.538016,
            "package-0": 49.718928,
            "psys": 79.821329,
            "uncore": 28.878832
          },
          "logits": {
            "l2": 1150.3575819756547,
            "sha256": "2c23050136374b58161b79874f63495a8c895d72d93c7555b2aaf74e81ff149c",
            "sum": 369774.01646265644,
            "top1": 220
          },
          "net_j": {
            "core": 5.8278301039888625,
            "dram": 2.6337627298383044,
            "package-0": 39.23228633144328,
            "psys": 61.22313122126779,
            "uncore": 28.875255571898045
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.831177665,
          "watts": {
            "core": 6.95431538042487,
            "dram": 1.9320987076368692,
            "package-0": 27.151340337039333,
            "psys": 43.59016087060018,
            "uncore": 15.77063359387359
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 811.0208018948907,
            "sha256": "b95537717dfcf6809ff402f9c640cbc97a728f4a0d1eb86875f847d4d31b12aa",
            "sum": 202420.8223354963,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 5.148425,
            "dram": 1.982172,
            "package-0": 14.736108,
            "psys": 25.097104,
            "uncore": 5.590318
          },
          "net_j": {
            "core": 2.76387879429272,
            "dram": 1.6699800788383095,
            "package-0": 11.115612595592406,
            "psys": 18.676107486877914,
            "uncore": 5.589083244211008
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.63221101,
          "token_latencies_s": [
            0.049636774,
            0.045743687,
            0.04519147,
            0.044705648,
            0.045360017,
            0.044527445,
            0.044626217,
            0.044621231,
            0.044313822,
            0.04521814,
            0.044709359,
            0.044605963,
            0.044476419,
            0.044447921
          ],
          "watts": {
            "core": 8.143523156928254,
            "dram": 3.135301297584172,
            "package-0": 23.308844305005064,
            "psys": 39.69735357819852,
            "uncore": 8.842487573887711
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 18.014907,
            "dram": 5.691758,
            "package-0": 65.374405,
            "psys": 106.951265,
            "uncore": 35.025301
          },
          "net_j": {
            "core": 8.56128840175079,
            "dram": 4.454061978403341,
            "package-0": 51.02082166071997,
            "psys": 81.4949949933806,
            "uncore": 35.020405766574385
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7142857142857143,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.043757,
            "dram": 0.15979,
            "package-0": 2.702141,
            "psys": 3.737112,
            "uncore": 0.319336
          },
          "logits": {
            "l2": 979.727719530536,
            "sha256": "8eca8a515a86885daba895c0e4490a5dfd1090aa41c110d1f8b0c9ea55c06524",
            "sum": 284299.187461914,
            "top1": 78
          },
          "net_j": {
            "core": 1.7975150970902796,
            "dram": 0.12755127358912122,
            "package-0": 2.3282679017461585,
            "psys": 3.0740431124340803,
            "uncore": 0.3192084920429798
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.065285731,
          "watts": {
            "core": 31.304803801614778,
            "dram": 2.4475486075203783,
            "package-0": 41.38945767490909,
            "psys": 57.24240109986668,
            "uncore": 4.891359798054494
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1634.2013256294047,
          "decode_tokens_per_s": 21.51216734208275,
          "end_to_end_ms": 2506.423129,
          "handoff_estimate_ms": 20.579845,
          "handoff_plus_first_decode_ms": 65.285731,
          "output_tokens_per_s_e2e": 6.383598928239862,
          "prefill_tokens_per_s": 2255.211050258657,
          "steady_tpot_ms": 44.705886,
          "steady_tpot_p95_ms": 47.207606299999995,
          "ttft_ms": 1809.143317
        },
        "prefill": {
          "gpu_active_fraction": 0.8146067415730337,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 178,
          "gross_j": {
            "core": 10.683444,
            "dram": 3.541617,
            "package-0": 47.787048,
            "psys": 77.50145,
            "uncore": 29.090379
          },
          "logits": {
            "l2": 1150.3575819756547,
            "sha256": "2c23050136374b58161b79874f63495a8c895d72d93c7555b2aaf74e81ff149c",
            "sum": 369774.01646265644,
            "top1": 220
          },
          "net_j": {
            "core": 3.8597953069267303,
            "dram": 2.6482445043342517,
            "package-0": 37.426590871749084,
            "psys": 59.12704217341829,
            "uncore": 29.086845606641695
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.809143317,
          "watts": {
            "core": 5.905250236181261,
            "dram": 1.9576210279862534,
            "package-0": 26.414185958049227,
            "psys": 42.838756483105094,
            "uncore": 16.079643180640286
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 811.0208018948907,
            "sha256": "b95537717dfcf6809ff402f9c640cbc97a728f4a0d1eb86875f847d4d31b12aa",
            "sum": 202420.8223354963,
            "top1": 42
          },
          "gpu_active_fraction": 0.9838709677419355,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 5.287706,
            "dram": 1.990351,
            "package-0": 14.885216,
            "psys": 25.712703,
            "uncore": 5.615586
          },
          "net_j": {
            "core": 2.9039779977337803,
            "dram": 1.6782662004799678,
            "package-0": 11.265962887224724,
            "psys": 19.293909707528236,
            "uncore": 5.614351667889709
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.631994081,
          "token_latencies_s": [
            0.048192979,
            0.046677021,
            0.044863513,
            0.044712599,
            0.04520041,
            0.044936164,
            0.044699173,
            0.044697623,
            0.044491113,
            0.04505052,
            0.04463654,
            0.0445817,
            0.044680747,
            0.044545521
          ],
          "watts": {
            "core": 8.366701776119957,
            "dram": 3.1493190519295386,
            "package-0": 23.552777545712487,
            "psys": 40.68503768154753,
            "uncore": 8.885504103320867
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.203982960432768,
    "compile_target": "NPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 397910,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 4096,
      "NPUW_LLM_MIN_RESPONSE_LEN": 4096,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04610307049006224,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 21.868902,
            "dram": 6.653792,
            "package-0": 66.56519899999999,
            "psys": 114.673047,
            "uncore": 27.412283
          },
          "net_j": {
            "core": 11.143050800064472,
            "dram": 5.249531481085421,
            "package-0": 50.27996414701965,
            "psys": 85.79096747048017,
            "uncore": 27.408172559420812
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.03571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 28,
          "gross_j": {
            "core": 3.917349,
            "dram": 0.332518,
            "package-0": 4.809131,
            "psys": 7.765117,
            "uncore": 0.008301
          },
          "logits": {
            "l2": 1147.84930727624,
            "sha256": "d711afeaae9e5c846617e5fbdf30231be31637936f41d7fa6839f09d6e1678c1",
            "sum": 302765.76277617365,
            "top1": 13
          },
          "net_j": {
            "core": 2.8615337185781744,
            "dram": 0.19428750553106047,
            "package-0": 3.206069434180092,
            "psys": 4.922066098203895,
            "uncore": 0.007754282126502436
          },
          "npu_busy_delta_us": 70916,
          "npu_utilization": 0.2533377885955956,
          "seconds": 0.279926656,
          "watts": {
            "core": 13.994197823018327,
            "dram": 1.1878754412012837,
            "package-0": 17.179968027053487,
            "psys": 27.739826963817265,
            "uncore": 0.029654196276327465
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1440.362704765708,
          "decode_tokens_per_s": 14.719507508358959,
          "end_to_end_ms": 2843.7281709999997,
          "handoff_estimate_ms": 228.932571,
          "handoff_plus_first_decode_ms": 279.926656,
          "output_tokens_per_s_e2e": 5.626416815491047,
          "prefill_tokens_per_s": 2236.0178960722633,
          "steady_tpot_ms": 50.994085,
          "steady_tpot_p95_ms": 60.5503181,
          "ttft_ms": 1824.672337
        },
        "prefill": {
          "gpu_active_fraction": 0.9888888888888889,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 180,
          "gross_j": {
            "core": 13.24545,
            "dram": 4.137441,
            "package-0": 50.295831,
            "psys": 85.70492,
            "uncore": 27.403982
          },
          "logits": {
            "l2": 3198.4374483838114,
            "sha256": "8ffcdb9ef1495d512db0d7eb0c62eab30d7c7aae955c1040aec2872d5443f1cc",
            "sum": 1194945.9496211372,
            "top1": 283
          },
          "net_j": {
            "core": 6.363229619414751,
            "dram": 3.2364001238238296,
            "package-0": 39.84644352309073,
            "psys": 67.17279302317412,
            "uncore": 27.40041827729431
          },
          "npu_busy_delta_us": 8610,
          "npu_utilization": 0.00471865541303485,
          "seconds": 1.824672337,
          "watts": {
            "core": 7.259084127826069,
            "dram": 2.26749806861351,
            "package-0": 27.564308385741697,
            "psys": 46.970033064078834,
            "uncore": 15.018577003833867
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1257.9593452313686,
            "sha256": "2bbcc434bbf921b754cb1245e19feaa54c0a23ec8ece6bc4bb3fdb60eebe175d",
            "sum": 352825.7396939341,
            "top1": 198
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 73,
          "gross_j": {
            "core": 4.706103,
            "dram": 2.183833,
            "package-0": 11.460237,
            "psys": 21.20301,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.918287462071547,
            "dram": 1.8188438517305303,
            "package-0": 7.227451189748827,
            "psys": 13.69610834910215,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 722397,
          "npu_utilization": 0.9773623089197,
          "seconds": 0.739129178,
          "token_latencies_s": [
            0.069330353,
            0.055822607,
            0.05318779,
            0.051765133,
            0.050952626,
            0.050992144,
            0.051133438,
            0.050996026,
            0.050801898,
            0.050654539,
            0.050718619,
            0.050753798,
            0.050704146,
            0.051280826
          ],
          "watts": {
            "core": 6.367091355714278,
            "dram": 2.9546026121025357,
            "package-0": 15.505052893474055,
            "psys": 28.686474071248206,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 22.45142,
            "dram": 6.555525,
            "package-0": 66.960827,
            "psys": 113.661268,
            "uncore": 27.36254
          },
          "net_j": {
            "core": 12.214156750693148,
            "dram": 5.2152318768935375,
            "package-0": 51.417423203036165,
            "psys": 86.09483566567036,
            "uncore": 27.358060290886876
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.07142857142857142,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 14,
          "gross_j": {
            "core": 2.605767,
            "dram": 0.229736,
            "package-0": 3.216911,
            "psys": 4.834765,
            "uncore": 0.024902
          },
          "logits": {
            "l2": 1290.0354581548856,
            "sha256": "0ba7257f7692ff17f9b89be8752da58a507ed8ad6c8b71db0317ad9ce49ea86e",
            "sum": 314371.7607283406,
            "top1": 13
          },
          "net_j": {
            "core": 2.0895980925147692,
            "dram": 0.1621576228286722,
            "package-0": 2.433203327246363,
            "psys": 3.4448490708340787,
            "uncore": 0.024634719565219912
          },
          "npu_busy_delta_us": 66808,
          "npu_utilization": 0.48818037618942456,
          "seconds": 0.136851056,
          "watts": {
            "core": 19.040898011046405,
            "dram": 1.67873019554924,
            "package-0": 23.506658216798854,
            "psys": 35.32866417925193,
            "uncore": 0.18196425170442238
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1509.1060637030785,
          "decode_tokens_per_s": 17.237192336887603,
          "end_to_end_ms": 2714.1896110000002,
          "handoff_estimate_ms": 85.8982135,
          "handoff_plus_first_decode_ms": 136.851056,
          "output_tokens_per_s_e2e": 5.89494556134015,
          "prefill_tokens_per_s": 2212.60740063433,
          "steady_tpot_ms": 50.952842499999996,
          "steady_tpot_p95_ms": 58.70467319999999,
          "ttft_ms": 1843.978285
        },
        "prefill": {
          "gpu_active_fraction": 0.994535519125683,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 183,
          "gross_j": {
            "core": 14.36648,
            "dram": 4.15038,
            "package-0": 51.473012,
            "psys": 86.677024,
            "uncore": 27.337027
          },
          "logits": {
            "l2": 3214.1518144681254,
            "sha256": "71fdd7f6aae8afe59eef177046b94bc6c134505127e15a4c245b1f55c8e9e449",
            "sum": 1201170.42534234,
            "top1": 283
          },
          "net_j": {
            "core": 7.4114422778582,
            "dram": 3.2398056596787375,
            "package-0": 40.91306476550008,
            "psys": 67.94881782994635,
            "uncore": 27.333425571321655
          },
          "npu_busy_delta_us": 6987,
          "npu_utilization": 0.0037890901735862905,
          "seconds": 1.843978285,
          "watts": {
            "core": 7.791024502221836,
            "dram": 2.2507748782952723,
            "package-0": 27.914109628465607,
            "psys": 47.00544724690183,
            "uncore": 14.825026532240319
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1164.0301254401054,
            "sha256": "8389285ef6e290edb74f8be63da61a7e17fe2670ddaa25ba68a3f64e2f9d5ebc",
            "sum": 214924.83903697878,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 72,
          "gross_j": {
            "core": 5.479173,
            "dram": 2.175409,
            "package-0": 12.270904,
            "psys": 22.149479,
            "uncore": 0.000611
          },
          "net_j": {
            "core": 2.7131163803201783,
            "dram": 1.8132685943861275,
            "package-0": 8.071155110289723,
            "psys": 14.701168764889932,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 717679,
          "npu_utilization": 0.9786172354278204,
          "seconds": 0.73336027,
          "token_latencies_s": [
            0.064988205,
            0.055321233,
            0.052661703,
            0.051518109,
            0.050851721,
            0.051035125,
            0.051041347,
            0.050948068,
            0.050688416,
            0.050782734,
            0.050854417,
            0.050957617,
            0.050938788,
            0.050744871
          ],
          "watts": {
            "core": 7.471325110099024,
            "dram": 2.966357858464299,
            "package-0": 16.732436296283133,
            "psys": 30.202725598974702,
            "uncore": 0.0008331512150228699
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 19.914745,
            "dram": 6.549727,
            "package-0": 64.392231,
            "psys": 110.26857700000001,
            "uncore": 27.417654
          },
          "net_j": {
            "core": 9.789597155078177,
            "dram": 5.224112360960237,
            "package-0": 49.019053850508044,
            "psys": 83.0040438777303,
            "uncore": 27.414089163615547
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.630853,
            "dram": 0.222106,
            "package-0": 3.200309,
            "psys": 4.749622,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1327.510740782216,
            "sha256": "a5cfa6d9c7d61c4d5314c5fb9e4394a7751d36c9ff305e9e941c701874e43519",
            "sum": 303319.145005431,
            "top1": 279
          },
          "net_j": {
            "core": 2.15504600199248,
            "dram": 0.15981192473544248,
            "package-0": 2.4778834728501735,
            "psys": 3.4683907670664413,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 66405,
          "npu_utilization": 0.5263972881292731,
          "seconds": 0.126149966,
          "watts": {
            "core": 20.854964003716024,
            "dram": 1.7606504943489243,
            "package-0": 25.369083333720436,
            "psys": 37.65060071439099,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1525.8163418326792,
          "decode_tokens_per_s": 17.45765204817119,
          "end_to_end_ms": 2684.4646289999996,
          "handoff_estimate_ms": 75.1438985,
          "handoff_plus_first_decode_ms": 126.149966,
          "output_tokens_per_s_e2e": 5.960220085283903,
          "prefill_tokens_per_s": 2235.319349104262,
          "steady_tpot_ms": 51.0060675,
          "steady_tpot_p95_ms": 58.60815495,
          "ttft_ms": 1825.242555
        },
        "prefill": {
          "gpu_active_fraction": 0.988950276243094,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 181,
          "gross_j": {
            "core": 12.878141,
            "dram": 4.142934,
            "package-0": 49.985224,
            "psys": 84.821499,
            "uncore": 27.417654
          },
          "logits": {
            "l2": 3223.526830340574,
            "sha256": "048b260b2c92587d73b6713099b0f5b80ad32e212c3094a362ae4de7628d322f",
            "sum": 1204873.0273589566,
            "top1": 283
          },
          "net_j": {
            "core": 5.993769895768977,
            "dram": 3.241611544651552,
            "package-0": 39.53257104429353,
            "psys": 66.28358065331216,
            "uncore": 27.414089163615547
          },
          "npu_busy_delta_us": 8724,
          "npu_utilization": 0.004779638725878819,
          "seconds": 1.825242555,
          "watts": {
            "core": 7.055577881811987,
            "dram": 2.269799150064195,
            "package-0": 27.385524111889886,
            "psys": 46.47135733694309,
            "uncore": 15.021375611089672
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1042.1375812788901,
            "sha256": "c93af3e216090c756e1d61880a00a0f36db91d207f91589a55d400f2ecd8f9d2",
            "sum": 156904.19420131668,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 73,
          "gross_j": {
            "core": 4.405751,
            "dram": 2.184687,
            "package-0": 11.206698,
            "psys": 20.697456,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.6407812573167195,
            "dram": 1.8226888915732418,
            "package-0": 7.008599333364334,
            "psys": 13.2520724573517,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 717462,
          "npu_utilization": 0.9787059037853887,
          "seconds": 0.733072108,
          "token_latencies_s": [
            0.062741529,
            0.056382492,
            0.052586517,
            0.051468474,
            0.051089063,
            0.051095787,
            0.050898616,
            0.050934025,
            0.050913882,
            0.050999066,
            0.050970306,
            0.050998315,
            0.050931558,
            0.051013069
          ],
          "watts": {
            "core": 6.009983127062311,
            "dram": 2.980180225326483,
            "package-0": 15.287306497821357,
            "psys": 28.233860999660347,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.1733186691999435,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 397758,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04595974553376436,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 18.286574,
            "dram": 5.473497,
            "package-0": 64.829669,
            "psys": 106.854402,
            "uncore": 34.603488
          },
          "net_j": {
            "core": 9.21583432980364,
            "dram": 4.285928632811989,
            "package-0": 51.05741698327603,
            "psys": 82.42913078723541,
            "uncore": 34.59879102734692
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.01562,
            "dram": 0.165527,
            "package-0": 2.741937,
            "psys": 3.932363,
            "uncore": 0.366637
          },
          "logits": {
            "l2": 979.727719530536,
            "sha256": "8eca8a515a86885daba895c0e4490a5dfd1090aa41c110d1f8b0c9ea55c06524",
            "sum": 284299.187461914,
            "top1": 78
          },
          "net_j": {
            "core": 1.742072504885852,
            "dram": 0.12971334378975105,
            "package-0": 2.3266053756525378,
            "psys": 3.195766867747554,
            "uncore": 0.3664953527749021
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.07252522,
          "watts": {
            "core": 27.791987394178193,
            "dram": 2.282337095978475,
            "package-0": 37.80666918349231,
            "psys": 54.220628355212156,
            "uncore": 5.055303520623585
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1703.181505254159,
          "decode_tokens_per_s": 21.299049867702585,
          "end_to_end_ms": 2404.911037,
          "handoff_estimate_ms": 27.866270999999998,
          "handoff_plus_first_decode_ms": 72.52522,
          "output_tokens_per_s_e2e": 6.653052754899059,
          "prefill_tokens_per_s": 2399.07668005914,
          "steady_tpot_ms": 44.65894900000001,
          "steady_tpot_p95_ms": 47.0527045,
          "ttft_ms": 1700.65427
        },
        "prefill": {
          "gpu_active_fraction": 0.8452380952380952,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 168,
          "gross_j": {
            "core": 10.665378,
            "dram": 3.325919,
            "package-0": 46.885927,
            "psys": 76.701342,
            "uncore": 28.630786
          },
          "logits": {
            "l2": 1150.3575819756547,
            "sha256": "2c23050136374b58161b79874f63495a8c895d72d93c7555b2aaf74e81ff149c",
            "sum": 369774.01646265644,
            "top1": 220
          },
          "net_j": {
            "core": 4.25092361277194,
            "dram": 2.4861194467098215,
            "package-0": 37.14675626170307,
            "psys": 59.42879380806612,
            "uncore": 28.62746449393361
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.70065427,
          "watts": {
            "core": 6.271338147994066,
            "dram": 1.9556702727121602,
            "package-0": 27.569346590356663,
            "psys": 45.101078657215844,
            "uncore": 16.835159564795024
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 811.0208018948907,
            "sha256": "b95537717dfcf6809ff402f9c640cbc97a728f4a0d1eb86875f847d4d31b12aa",
            "sum": 202420.8223354963,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 5.605576,
            "dram": 1.982051,
            "package-0": 15.201805,
            "psys": 26.220697,
            "uncore": 5.606065
          },
          "net_j": {
            "core": 3.222838212145847,
            "dram": 1.670095842312416,
            "package-0": 11.584055345920422,
            "psys": 19.804570111421736,
            "uncore": 5.604831180638413
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.631731547,
          "token_latencies_s": [
            0.049360224,
            0.045810194,
            0.044899891,
            0.044527344,
            0.0446105,
            0.044837909,
            0.044626674,
            0.04505305,
            0.044372493,
            0.044386252,
            0.044665544,
            0.044652354,
            0.04533809,
            0.044561265
          ],
          "watts": {
            "core": 8.873351388924702,
            "dram": 3.1374893487787148,
            "package-0": 24.06371040387508,
            "psys": 41.50607504804569,
            "uncore": 8.874125451898637
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 19.070263999999998,
            "dram": 5.676011,
            "package-0": 66.26600499999999,
            "psys": 110.32820899999999,
            "uncore": 34.765109
          },
          "net_j": {
            "core": 9.503069146906086,
            "dram": 4.423445235425221,
            "package-0": 51.739976980163604,
            "psys": 84.56610607029198,
            "uncore": 34.76015495499094
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.463556,
            "dram": 0.164978,
            "package-0": 3.143913,
            "psys": 4.352284,
            "uncore": 0.323791
          },
          "logits": {
            "l2": 979.727719530536,
            "sha256": "8eca8a515a86885daba895c0e4490a5dfd1090aa41c110d1f8b0c9ea55c06524",
            "sum": 284299.187461914,
            "top1": 78
          },
          "net_j": {
            "core": 2.182120867412432,
            "dram": 0.12813167070069112,
            "package-0": 2.7166054470017107,
            "psys": 3.594448399796637,
            "uncore": 0.3236452684313031
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.074616457,
          "watts": {
            "core": 33.01625538183889,
            "dram": 2.211013583772813,
            "package-0": 42.13431093357863,
            "psys": 58.32874107115539,
            "uncore": 4.339404643669962
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1614.8010239655146,
          "decode_tokens_per_s": 21.1610435402155,
          "end_to_end_ms": 2536.5354239999997,
          "handoff_estimate_ms": 29.993397499999997,
          "handoff_plus_first_decode_ms": 74.616457,
          "output_tokens_per_s_e2e": 6.307816499865291,
          "prefill_tokens_per_s": 2232.331310243325,
          "steady_tpot_ms": 44.6230595,
          "steady_tpot_p95_ms": 48.2103175,
          "ttft_ms": 1827.685694
        },
        "prefill": {
          "gpu_active_fraction": 0.8066298342541437,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 181,
          "gross_j": {
            "core": 10.954623,
            "dram": 3.527274,
            "package-0": 47.863586,
            "psys": 79.894815,
            "uncore": 28.836047
          },
          "logits": {
            "l2": 1150.3575819756547,
            "sha256": "2c23050136374b58161b79874f63495a8c895d72d93c7555b2aaf74e81ff149c",
            "sum": 369774.01646265644,
            "top1": 220
          },
          "net_j": {
            "core": 4.061036969517589,
            "dram": 2.624745098945908,
            "package-0": 37.39694187143839,
            "psys": 61.33208312309528,
            "uncore": 28.8324773919808
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.827685694,
          "watts": {
            "core": 5.993712724218544,
            "dram": 1.9299127916684344,
            "package-0": 26.188083737334324,
            "psys": 43.71365123789167,
            "uncore": 15.77735553474218
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 811.0208018948907,
            "sha256": "b95537717dfcf6809ff402f9c640cbc97a728f4a0d1eb86875f847d4d31b12aa",
            "sum": 202420.8223354963,
            "top1": 42
          },
          "gpu_active_fraction": 0.9841269841269841,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 5.652085,
            "dram": 1.983759,
            "package-0": 15.258506,
            "psys": 26.08111,
            "uncore": 5.605271
          },
          "net_j": {
            "core": 3.259911309976064,
            "dram": 1.6705684657786222,
            "package-0": 11.626429661723503,
            "psys": 19.639574547400066,
            "uncore": 5.604032294578834
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.634233273,
          "token_latencies_s": [
            0.05384143,
            0.04456289,
            0.044729934,
            0.044597111,
            0.044737294,
            0.044448042,
            0.044603898,
            0.044754349,
            0.044638263,
            0.044617655,
            0.044386376,
            0.044628464,
            0.044482925,
            0.04517818
          ],
          "watts": {
            "core": 8.911681617183145,
            "dram": 3.1278065728349134,
            "package-0": 24.058192229217845,
            "psys": 41.12226701168357,
            "uncore": 8.837869658724134
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 19.28236,
            "dram": 5.686386,
            "package-0": 66.565076,
            "psys": 109.25796,
            "uncore": 34.852450000000005
          },
          "net_j": {
            "core": 9.78819250742739,
            "dram": 4.443381195881322,
            "package-0": 52.149926612974575,
            "psys": 83.69250179294299,
            "uncore": 34.84753376971474
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 7,
          "gross_j": {
            "core": 2.067866,
            "dram": 0.165405,
            "package-0": 2.747552,
            "psys": 3.844228,
            "uncore": 0.329223
          },
          "logits": {
            "l2": 979.727719530536,
            "sha256": "8eca8a515a86885daba895c0e4490a5dfd1090aa41c110d1f8b0c9ea55c06524",
            "sum": 284299.187461914,
            "top1": 78
          },
          "net_j": {
            "core": 1.806476744223326,
            "dram": 0.13118313684774463,
            "package-0": 2.350680428782,
            "psys": 3.140371016271821,
            "uncore": 0.3290876485037791
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.069301725,
          "watts": {
            "core": 29.838593483784138,
            "dram": 2.3867371266732538,
            "package-0": 39.646228142228786,
            "psys": 55.470884743489435,
            "uncore": 4.7505743904643065
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1627.221771402249,
          "decode_tokens_per_s": 21.359745537762972,
          "end_to_end_ms": 2517.1737940000003,
          "handoff_estimate_ms": 24.61765149999999,
          "handoff_plus_first_decode_ms": 69.30172499999999,
          "output_tokens_per_s_e2e": 6.356335044540035,
          "prefill_tokens_per_s": 2248.0351548751337,
          "steady_tpot_ms": 44.684073500000004,
          "steady_tpot_p95_ms": 48.42182755,
          "ttft_ms": 1814.918237
        },
        "prefill": {
          "gpu_active_fraction": 0.8166666666666667,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 180,
          "gross_j": {
            "core": 11.786407,
            "dram": 3.532278,
            "package-0": 48.796445,
            "psys": 79.77659,
            "uncore": 28.936205
          },
          "logits": {
            "l2": 1150.3575819756547,
            "sha256": "2c23050136374b58161b79874f63495a8c895d72d93c7555b2aaf74e81ff149c",
            "sum": 369774.01646265644,
            "top1": 220
          },
          "net_j": {
            "core": 4.940976712875935,
            "dram": 2.6360537928948147,
            "package-0": 38.40291652216784,
            "psys": 61.34352969727182,
            "uncore": 28.932660327787346
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.814918237,
          "watts": {
            "core": 6.494180707271167,
            "dram": 1.946246353135301,
            "package-0": 26.886304851208564,
            "psys": 43.95602422942648,
            "uncore": 15.943530904086673
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 811.0208018948907,
            "sha256": "b95537717dfcf6809ff402f9c640cbc97a728f4a0d1eb86875f847d4d31b12aa",
            "sum": 202420.8223354963,
            "top1": 42
          },
          "gpu_active_fraction": 0.9838709677419355,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 5.428087,
            "dram": 1.988703,
            "package-0": 15.021079,
            "psys": 25.637142,
            "uncore": 5.587022
          },
          "net_j": {
            "core": 3.0407390503281286,
            "dram": 1.676144266138763,
            "package-0": 11.396329662024732,
            "psys": 19.208601079399354,
            "uncore": 5.5857857934236135
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.632953832,
          "token_latencies_s": [
            0.049069171,
            0.048073258,
            0.044983623,
            0.044807323,
            0.044667437,
            0.044410077,
            0.044877966,
            0.044680398,
            0.044728233,
            0.044556619,
            0.044687749,
            0.044327617,
            0.044534228,
            0.044527339
          ],
          "watts": {
            "core": 8.575802413342526,
            "dram": 3.141940058591825,
            "package-0": 23.731713500393187,
            "psys": 40.50396838422174,
            "uncore": 8.826902875911493
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 1.6647716462612152,
    "compile_target": "AUTO:GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 398324,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.047261254861950874,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 35.882416,
            "dram": 10.622043,
            "package-0": 137.941174,
            "psys": 223.042153,
            "uncore": 78.186018
          },
          "net_j": {
            "core": 16.498080529450466,
            "dram": 8.084187823820436,
            "package-0": 108.50962122880789,
            "psys": 170.84490466499898,
            "uncore": 78.17598048410674
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 9,
          "gross_j": {
            "core": 2.396784,
            "dram": 0.181091,
            "package-0": 3.173209,
            "psys": 4.477527,
            "uncore": 0.386962
          },
          "logits": {
            "l2": 1104.323434566963,
            "sha256": "4cc0c5b7c789d0230c86bfae04035aec63fca0279550b97617a206b017f84bae",
            "sum": 343667.50144172134,
            "top1": 261
          },
          "net_j": {
            "core": 2.078594251096151,
            "dram": 0.13943264640362485,
            "package-0": 2.690096313452458,
            "psys": 3.6207202569254595,
            "uncore": 0.3867972363088975
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.084361151,
          "watts": {
            "core": 28.410992163916777,
            "dram": 2.1466160413102946,
            "package-0": 37.61457688029885,
            "psys": 53.07569831521147,
            "uncore": 4.586969184429454
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1593.9794344485392,
          "decode_tokens_per_s": 19.545877697122307,
          "end_to_end_ms": 5139.338578,
          "handoff_estimate_ms": 36.083645999999995,
          "handoff_plus_first_decode_ms": 84.36115099999999,
          "output_tokens_per_s_e2e": 3.113241082907303,
          "prefill_tokens_per_s": 1870.1194143434132,
          "steady_tpot_ms": 48.277505,
          "steady_tpot_p95_ms": 50.668234749999996,
          "ttft_ms": 4371.913332
        },
        "prefill": {
          "gpu_active_fraction": 0.8752886836027713,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 433,
          "gross_j": {
            "core": 27.451346,
            "dram": 8.360452,
            "package-0": 118.330446,
            "psys": 190.899841,
            "uncore": 71.609802
          },
          "logits": {
            "l2": 1063.3971159220778,
            "sha256": "084ff5e6e0b62105635ea609877c5eba14bc7d2005fb94d197b9e77e6987b4d0",
            "sum": 342411.9841043374,
            "top1": 220
          },
          "net_j": {
            "core": 10.961551955600694,
            "dram": 6.201558844581072,
            "package-0": 93.29372269621841,
            "psys": 146.49688088634434,
            "uncore": 71.60126332349759
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.371913332,
          "watts": {
            "core": 6.279023373832974,
            "dram": 1.9123096377062399,
            "package-0": 27.066054840082543,
            "psys": 43.66505612147391,
            "uncore": 16.37951088276514
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1116.0949352903115,
            "sha256": "0b8aeac4667d48fb5fafee7dd6a8d3dfdcbf32877fc72e9ba5308bdb34b0f5fd",
            "sum": 350539.43440529774,
            "top1": 42
          },
          "gpu_active_fraction": 0.9852941176470589,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 6.034286,
            "dram": 2.0805,
            "package-0": 16.437519,
            "psys": 27.664785,
            "uncore": 6.189254
          },
          "net_j": {
            "core": 3.457934322753622,
            "dram": 1.7431963328357385,
            "package-0": 12.525802219137024,
            "psys": 20.72730352172917,
            "uncore": 6.187919924300253
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.683064095,
          "token_latencies_s": [
            0.053622345,
            0.048928247,
            0.048392344,
            0.04821759,
            0.04907756,
            0.048030054,
            0.04819754,
            0.048201542,
            0.048895742,
            0.048197996,
            0.048193964,
            0.048271342,
            0.048522036,
            0.048283668
          ],
          "watts": {
            "core": 8.834143156067952,
            "dram": 3.0458342273136165,
            "package-0": 24.064387398374382,
            "psys": 40.50100891337291,
            "uncore": 9.061014984252687
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 38.53457,
            "dram": 11.048067,
            "package-0": 142.17786,
            "psys": 231.643511,
            "uncore": 78.781109
          },
          "net_j": {
            "core": 18.291663057351624,
            "dram": 8.397805078802191,
            "package-0": 111.44272415641893,
            "psys": 177.1343408345819,
            "uncore": 78.77062690220143
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.283198,
            "dram": 0.174743,
            "package-0": 3.013787,
            "psys": 4.228627,
            "uncore": 0.354919
          },
          "logits": {
            "l2": 1104.323434566963,
            "sha256": "4cc0c5b7c789d0230c86bfae04035aec63fca0279550b97617a206b017f84bae",
            "sum": 343667.50144172134,
            "top1": 261
          },
          "net_j": {
            "core": 1.981982111162069,
            "dram": 0.13530691493036826,
            "package-0": 2.5564460020935202,
            "psys": 3.4175266877912667,
            "uncore": 0.354763025642452
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.079860898,
          "watts": {
            "core": 28.589686031329126,
            "dram": 2.188092099840901,
            "package-0": 37.7379553132498,
            "psys": 52.949905471886886,
            "uncore": 4.444214989919097
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1526.3732712918795,
          "decode_tokens_per_s": 19.532305787252795,
          "end_to_end_ms": 5366.970291,
          "handoff_estimate_ms": 30.9977855,
          "handoff_plus_first_decode_ms": 79.860898,
          "output_tokens_per_s_e2e": 2.981197795491952,
          "prefill_tokens_per_s": 1777.773214865182,
          "steady_tpot_ms": 48.8631125,
          "steady_tpot_p95_ms": 50.528452200000004,
          "ttft_ms": 4599.011804
        },
        "prefill": {
          "gpu_active_fraction": 0.8549450549450549,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 455,
          "gross_j": {
            "core": 29.746994,
            "dram": 8.780312,
            "package-0": 122.208977,
            "psys": 198.94053,
            "uncore": 72.219846
          },
          "logits": {
            "l2": 1063.3971159220778,
            "sha256": "084ff5e6e0b62105635ea609877c5eba14bc7d2005fb94d197b9e77e6987b4d0",
            "sum": 342411.9841043374,
            "top1": 220
          },
          "net_j": {
            "core": 12.400639738894693,
            "dram": 6.5092754191342985,
            "package-0": 95.87172434570302,
            "psys": 152.23106387208279,
            "uncore": 72.21086378305218
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.599011804,
          "watts": {
            "core": 6.46812734294952,
            "dram": 1.9091736168981577,
            "package-0": 26.57287743721564,
            "psys": 43.25723404905616,
            "uncore": 15.703339995167363
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1116.0949352903115,
            "sha256": "0b8aeac4667d48fb5fafee7dd6a8d3dfdcbf32877fc72e9ba5308bdb34b0f5fd",
            "sum": 350539.43440529774,
            "top1": 42
          },
          "gpu_active_fraction": 0.9855072463768116,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 69,
          "gross_j": {
            "core": 6.504378,
            "dram": 2.093012,
            "package-0": 16.955096,
            "psys": 28.474354,
            "uncore": 6.206344
          },
          "net_j": {
            "core": 3.9090412072948606,
            "dram": 1.7532227447375246,
            "package-0": 13.014553808622402,
            "psys": 21.485750274707847,
            "uncore": 6.205000093506804
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.688097589,
          "token_latencies_s": [
            0.053016543,
            0.049174865,
            0.048851492,
            0.049009647,
            0.049086684,
            0.049188711,
            0.048749696,
            0.04889204,
            0.048874733,
            0.048674549,
            0.048647884,
            0.048798816,
            0.048627863,
            0.048472511
          ],
          "watts": {
            "core": 9.452696977841033,
            "dram": 3.0417371510365805,
            "package-0": 24.64053975925209,
            "psys": 41.381272736882096,
            "uncore": 9.019569461098635
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 38.108239999999995,
            "dram": 11.046907000000001,
            "package-0": 142.011294,
            "psys": 234.02643899999998,
            "uncore": 79.02335699999999
          },
          "net_j": {
            "core": 17.845775954967014,
            "dram": 8.394084604464897,
            "package-0": 111.24646428882369,
            "psys": 179.464606367435,
            "uncore": 79.01286477522409
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 9,
          "gross_j": {
            "core": 2.442743,
            "dram": 0.175353,
            "package-0": 3.163383,
            "psys": 4.446522,
            "uncore": 0.345275
          },
          "logits": {
            "l2": 1104.323434566963,
            "sha256": "4cc0c5b7c789d0230c86bfae04035aec63fca0279550b97617a206b017f84bae",
            "sum": 343667.50144172134,
            "top1": 261
          },
          "net_j": {
            "core": 2.143483830722662,
            "dram": 0.13617309450757004,
            "package-0": 2.709012921323303,
            "psys": 3.6406906523622418,
            "uncore": 0.3451200388628287
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.079342116,
          "watts": {
            "core": 30.787469797251184,
            "dram": 2.210087263112569,
            "package-0": 39.87016176881393,
            "psys": 56.04239241615386,
            "uncore": 4.351724120894381
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1524.9000329790347,
          "decode_tokens_per_s": 19.58324790903004,
          "end_to_end_ms": 5372.155435,
          "handoff_estimate_ms": 30.770626000000004,
          "handoff_plus_first_decode_ms": 79.342116,
          "output_tokens_per_s_e2e": 2.978320376912177,
          "prefill_tokens_per_s": 1775.0009772871852,
          "steady_tpot_ms": 48.571490000000004,
          "steady_tpot_p95_ms": 50.694644,
          "ttft_ms": 4606.194646999999
        },
        "prefill": {
          "gpu_active_fraction": 0.8549450549450549,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 455,
          "gross_j": {
            "core": 29.327989,
            "dram": 8.78489,
            "package-0": 122.083733,
            "psys": 201.072117,
            "uncore": 72.471189
          },
          "logits": {
            "l2": 1063.3971159220778,
            "sha256": "084ff5e6e0b62105635ea609877c5eba14bc7d2005fb94d197b9e77e6987b4d0",
            "sum": 342411.9841043374,
            "top1": 220
          },
          "net_j": {
            "core": 11.95454280162862,
            "dram": 6.510306461678477,
            "package-0": 95.70534621538775,
            "psys": 154.28969894760223,
            "uncore": 72.462192754418
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.606194647,
          "watts": {
            "core": 6.367075481515143,
            "dram": 1.90719035412921,
            "package-0": 26.504249680267584,
            "psys": 43.65254454258846,
            "uncore": 15.733418701096415
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1116.0949352903115,
            "sha256": "0b8aeac4667d48fb5fafee7dd6a8d3dfdcbf32877fc72e9ba5308bdb34b0f5fd",
            "sum": 350539.43440529774,
            "top1": 42
          },
          "gpu_active_fraction": 0.9852941176470589,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 6.337508,
            "dram": 2.086664,
            "package-0": 16.764178,
            "psys": 28.5078,
            "uncore": 6.206893
          },
          "net_j": {
            "core": 3.7477493226157335,
            "dram": 1.7476050482788512,
            "package-0": 12.83210515211264,
            "psys": 21.534216767470532,
            "uncore": 6.205551981943257
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.686618672,
          "token_latencies_s": [
            0.052987064,
            0.049460264,
            0.048902616,
            0.049254507,
            0.048496768,
            0.048907001,
            0.049327183,
            0.048546731,
            0.048442555,
            0.048596249,
            0.048410434,
            0.048469362,
            0.048395614,
            0.048400413
          ],
          "watts": {
            "core": 9.230025716515906,
            "dram": 3.0390434823479424,
            "package-0": 24.415557985291727,
            "psys": 41.519115576862724,
            "uncore": 9.039796400992719
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 6.948272679001093,
    "compile_target": "NPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 398393,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 8192,
      "NPUW_LLM_MIN_RESPONSE_LEN": 8192,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04698196332901716,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 42.518811,
            "dram": 12.945523,
            "package-0": 141.026984,
            "psys": 243.361804,
            "uncore": 65.918471
          },
          "net_j": {
            "core": 21.012227852029998,
            "dram": 10.129816810487558,
            "package-0": 108.37318795890384,
            "psys": 185.44986458286223,
            "uncore": 65.90890867399104
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 30,
          "gross_j": {
            "core": 4.55376,
            "dram": 0.391235,
            "package-0": 5.577196,
            "psys": 8.942543,
            "uncore": 0.002624
          },
          "logits": {
            "l2": 1241.46397566741,
            "sha256": "bdb717b7c1752b3326560db2337bf6d62440ead312748b3fc50ce706e9507034",
            "sum": 266600.9037211053,
            "top1": 45195
          },
          "net_j": {
            "core": 3.371212115696847,
            "dram": 0.23641229411882345,
            "package-0": 3.7817142557737102,
            "psys": 5.758232365323812,
            "uncore": 0.002011657977686571
          },
          "npu_busy_delta_us": 76496,
          "npu_utilization": 0.2439853096832885,
          "seconds": 0.313527073,
          "watts": {
            "core": 14.52429596087863,
            "dram": 1.24785077172586,
            "package-0": 17.788562712094723,
            "psys": 28.522394938442847,
            "uncore": 0.008369293199761412
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1436.6871705245176,
          "decode_tokens_per_s": 13.398854565866928,
          "end_to_end_ms": 5702.006789,
          "handoff_estimate_ms": 257.9817665,
          "handoff_plus_first_decode_ms": 313.527073,
          "output_tokens_per_s_e2e": 2.8060296299306984,
          "prefill_tokens_per_s": 1784.1757844230365,
          "steady_tpot_ms": 55.5453065,
          "steady_tpot_p95_ms": 66.3210663,
          "ttft_ms": 4582.508109
        },
        "prefill": {
          "gpu_active_fraction": 0.9933774834437086,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 453,
          "gross_j": {
            "core": 33.703588,
            "dram": 10.205967,
            "package-0": 123.838489,
            "psys": 212.656621,
            "uncore": 65.915847
          },
          "logits": {
            "l2": 3222.6951479846207,
            "sha256": "73a80de707a59e79c2b90653ba89bb9f702fb9496328d98e51493e37449336cc",
            "sum": 1204541.814432323,
            "top1": 283
          },
          "net_j": {
            "core": 16.41948166351151,
            "dram": 7.943080103601698,
            "package-0": 97.59574838512992,
            "psys": 166.1147732247727,
            "uncore": 65.90689701601335
          },
          "npu_busy_delta_us": 8560,
          "npu_utilization": 0.0018679726901493629,
          "seconds": 4.582508109,
          "watts": {
            "core": 7.354834339257686,
            "dram": 2.227157433710937,
            "package-0": 27.024172364645125,
            "psys": 46.40616359900041,
            "uncore": 14.384229210755118
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1072.8463720276775,
            "sha256": "1b3c3c5e744d021148f21d2526bb88f65dc9872c0e17543459259b67a913c12f",
            "sum": 147567.20988509804,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 79,
          "gross_j": {
            "core": 4.261463,
            "dram": 2.348321,
            "package-0": 11.611299,
            "psys": 21.76264,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.2215340728216382,
            "dram": 1.9503244127670363,
            "package-0": 6.995725318000215,
            "psys": 13.576858992765713,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 785168,
          "npu_utilization": 0.9741881639262262,
          "seconds": 0.805971607,
          "token_latencies_s": [
            0.077868211,
            0.060103373,
            0.057391334,
            0.055917078,
            0.055396267,
            0.055286741,
            0.055271382,
            0.055393308,
            0.05590076,
            0.055726225,
            0.055651875,
            0.055282687,
            0.055438738,
            0.055307245
          ],
          "watts": {
            "core": 5.287361196087395,
            "dram": 2.9136522671573464,
            "package-0": 14.40658566524416,
            "psys": 27.00174523641749,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 40.231648,
            "dram": 12.756132000000001,
            "package-0": 138.361402,
            "psys": 239.398252,
            "uncore": 65.90168700000001
          },
          "net_j": {
            "core": 19.430646481315193,
            "dram": 10.032802667709165,
            "package-0": 106.77890205714984,
            "psys": 183.3862703905657,
            "uncore": 65.89247136745512
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.058823529411764705,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 17,
          "gross_j": {
            "core": 2.967094,
            "dram": 0.284301,
            "package-0": 3.702078,
            "psys": 5.760239,
            "uncore": 0.015015
          },
          "logits": {
            "l2": 1528.6449015081357,
            "sha256": "36c32280eaaaedd81d9ef77d3977727da67f2c40dd6aecc130b0cea07b2b713d",
            "sum": 357628.61823196337,
            "top1": 311
          },
          "net_j": {
            "core": 2.323819259206406,
            "dram": 0.20008154697263714,
            "package-0": 2.725383466993919,
            "psys": 4.0280583202005715,
            "uncore": 0.014681902151778139
          },
          "npu_busy_delta_us": 72848,
          "npu_utilization": 0.42713466791848015,
          "seconds": 0.170550427,
          "watts": {
            "core": 17.397165472942497,
            "dram": 1.6669615256958579,
            "package-0": 21.706647500800454,
            "psys": 33.77440386003842,
            "uncore": 0.08803847791011395
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1485.4204045296813,
          "decode_tokens_per_s": 15.512472556393572,
          "end_to_end_ms": 5514.937034,
          "handoff_estimate_ms": 114.75934750000002,
          "handoff_plus_first_decode_ms": 170.550427,
          "output_tokens_per_s_e2e": 2.9012117275970337,
          "prefill_tokens_per_s": 1797.7238755916362,
          "steady_tpot_ms": 55.791079499999995,
          "steady_tpot_p95_ms": 62.431346199999986,
          "ttft_ms": 4547.973196
        },
        "prefill": {
          "gpu_active_fraction": 0.9933184855233853,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 449,
          "gross_j": {
            "core": 32.496377,
            "dram": 10.134923,
            "package-0": 122.486442,
            "psys": 210.835337,
            "uncore": 65.886672
          },
          "logits": {
            "l2": 3232.9150534350383,
            "sha256": "688fa4cf4457f077fd4509aff964fb4016666e8871d4f91399f7169f143c3ec9",
            "sum": 1208575.6573289484,
            "top1": 283
          },
          "net_j": {
            "core": 15.342527958079316,
            "dram": 7.889089778191488,
            "package-0": 96.44147316042393,
            "psys": 164.64424008555568,
            "uncore": 65.87778946530334
          },
          "npu_busy_delta_us": 6840,
          "npu_utilization": 0.0015039666473883062,
          "seconds": 4.547973196,
          "watts": {
            "core": 7.14524373815153,
            "dram": 2.2284482698609116,
            "package-0": 26.932094082640674,
            "psys": 46.3580869793675,
            "uncore": 14.487040525645174
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1208.9667946533887,
            "sha256": "47f7337a58159a23820f7ac08b148be4ffd7e2a11bf4a8a77c0eaa9d27902624",
            "sum": 253333.13155601174,
            "top1": 311
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 79,
          "gross_j": {
            "core": 4.768177,
            "dram": 2.336908,
            "package-0": 12.172882,
            "psys": 22.802676,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.7642992640294697,
            "dram": 1.943631342545039,
            "package-0": 7.612045429731987,
            "psys": 14.713971984809458,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 779709,
          "npu_utilization": 0.9790254523978628,
          "seconds": 0.796413411,
          "token_latencies_s": [
            0.068014997,
            0.059424765,
            0.05689953,
            0.055778549,
            0.055316883,
            0.056162077,
            0.055753213,
            0.055873058,
            0.05580361,
            0.055914315,
            0.055476847,
            0.055436271,
            0.055292128,
            0.055237246
          ],
          "watts": {
            "core": 5.987062666376922,
            "dram": 2.93429011581524,
            "package-0": 15.284627094256704,
            "psys": 28.63170770990445,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 42.356764999999996,
            "dram": 12.727018000000001,
            "package-0": 140.656135,
            "psys": 243.387622,
            "uncore": 66.045424
          },
          "net_j": {
            "core": 21.611620737261823,
            "dram": 10.01100166672049,
            "package-0": 109.15844404008067,
            "psys": 187.5260502441502,
            "uncore": 66.03624405269
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.058823529411764705,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 17,
          "gross_j": {
            "core": 3.010002,
            "dram": 0.284668,
            "package-0": 3.750662,
            "psys": 5.88353,
            "uncore": 0.014648
          },
          "logits": {
            "l2": 1558.9311317044221,
            "sha256": "a8d46c1c7b08272c8d75f1a3412e0a4d9699cbd3f7abd50b4eeebcbeea7b6ffd",
            "sum": 375173.26933249086,
            "top1": 1753
          },
          "net_j": {
            "core": 2.3694154959204443,
            "dram": 0.20080049896345376,
            "package-0": 2.7780490606513792,
            "psys": 4.158588080567628,
            "uncore": 0.01431629416332184
          },
          "npu_busy_delta_us": 73085,
          "npu_utilization": 0.4303225987535311,
          "seconds": 0.169837699,
          "watts": {
            "core": 17.722814296960063,
            "dram": 1.676117856495453,
            "package-0": 22.0838013119808,
            "psys": 34.64207319483291,
            "uncore": 0.08624704695274986
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1489.419967351387,
          "decode_tokens_per_s": 15.468493217695292,
          "end_to_end_ms": 5500.127687,
          "handoff_estimate_ms": 114.17112300000001,
          "handoff_plus_first_decode_ms": 169.83769900000001,
          "output_tokens_per_s_e2e": 2.9090233737331777,
          "prefill_tokens_per_s": 1804.691333085977,
          "steady_tpot_ms": 55.666576,
          "steady_tpot_p95_ms": 63.613831950000005,
          "ttft_ms": 4530.4146200000005
        },
        "prefill": {
          "gpu_active_fraction": 0.9955357142857143,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 448,
          "gross_j": {
            "core": 34.339451,
            "dram": 10.101414,
            "package-0": 124.467272,
            "psys": 214.224732,
            "uncore": 66.030776
          },
          "logits": {
            "l2": 3237.686448552048,
            "sha256": "e70dbb184774b1dcb7c98aa69c0e3c5953b75cf3f79612dccf368f3a8f764feb",
            "sum": 1210447.6801931784,
            "top1": 283
          },
          "net_j": {
            "core": 17.25182863733063,
            "dram": 7.864251373115647,
            "package-0": 98.52285620791979,
            "psys": 168.21196724600222,
            "uncore": 66.02192775852667
          },
          "npu_busy_delta_us": 8523,
          "npu_utilization": 0.0018812847641746307,
          "seconds": 4.53041462,
          "watts": {
            "core": 7.579759002278691,
            "dram": 2.22968863719586,
            "package-0": 27.473704382492034,
            "psys": 47.2858998499347,
            "uncore": 14.574996228490892
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1151.2441687220637,
            "sha256": "53e96ef399ab02db12c66d59d43548dc5b1ae6a927dc046d7c6ddcad0edfdfbb",
            "sum": 109909.59551849216,
            "top1": 437
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 79,
          "gross_j": {
            "core": 5.007312,
            "dram": 2.340936,
            "package-0": 12.438201,
            "psys": 23.27936,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.9903766040107493,
            "dram": 1.9459497946413902,
            "package-0": 7.857538771509494,
            "psys": 15.15549491758037,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 782538,
          "npu_utilization": 0.9783249132382333,
          "seconds": 0.799875368,
          "token_latencies_s": [
            0.068475622,
            0.060995945,
            0.056762271,
            0.055456332,
            0.055669597,
            0.055801353,
            0.055663555,
            0.055581298,
            0.055382226,
            0.057817981,
            0.055618211,
            0.05545476,
            0.055486474,
            0.055681979
          ],
          "watts": {
            "core": 6.2601152633568775,
            "dram": 2.9266259390550453,
            "package-0": 15.550173811578105,
            "psys": 29.10373407073088,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.1653368724510074,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 398238,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_1_5b_ov_fp16/openvino_model.xml",
      "key": "qwen25_1_5b_fp16",
      "size_b": 1.5
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.046355665661394596,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 36.086394999999996,
            "dram": 10.645845000000001,
            "package-0": 138.196423,
            "psys": 224.759678,
            "uncore": 78.139143
          },
          "net_j": {
            "core": 16.602067007954226,
            "dram": 8.09489850383972,
            "package-0": 108.61304995219268,
            "psys": 172.29317439524198,
            "uncore": 78.12905370639554
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.6666666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 9,
          "gross_j": {
            "core": 2.50982,
            "dram": 0.182067,
            "package-0": 3.293937,
            "psys": 4.62804,
            "uncore": 0.386229
          },
          "logits": {
            "l2": 1104.323434566963,
            "sha256": "4cc0c5b7c789d0230c86bfae04035aec63fca0279550b97617a206b017f84bae",
            "sum": 343667.50144172134,
            "top1": 261
          },
          "net_j": {
            "core": 2.1798918196110164,
            "dram": 0.1388718158452579,
            "package-0": 2.793001661433737,
            "psys": 3.739624547706418,
            "uncore": 0.3860581579631749
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.087473343,
          "watts": {
            "core": 28.692398322995384,
            "dram": 2.0813998157129996,
            "package-0": 37.656466381992516,
            "psys": 52.908004213352186,
            "uncore": 4.415390869421785
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1585.7992178699267,
          "decode_tokens_per_s": 19.40710037826237,
          "end_to_end_ms": 5165.8494390000005,
          "handoff_estimate_ms": 38.863736499999995,
          "handoff_plus_first_decode_ms": 87.473343,
          "output_tokens_per_s_e2e": 3.0972640974022005,
          "prefill_tokens_per_s": 1861.1696519585068,
          "steady_tpot_ms": 48.6096065,
          "steady_tpot_p95_ms": 50.45135584999999,
          "ttft_ms": 4392.936448
        },
        "prefill": {
          "gpu_active_fraction": 0.8778801843317973,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 434,
          "gross_j": {
            "core": 26.490533,
            "dram": 8.377114,
            "package-0": 117.424565,
            "psys": 191.165648,
            "uncore": 71.575317
          },
          "logits": {
            "l2": 1063.3971159220778,
            "sha256": "084ff5e6e0b62105635ea609877c5eba14bc7d2005fb94d197b9e77e6987b4d0",
            "sum": 342411.9841043374,
            "top1": 220
          },
          "net_j": {
            "core": 9.921444873882717,
            "dram": 6.207839426242842,
            "package-0": 92.26744819838243,
            "psys": 146.5491684246523,
            "uncore": 71.5667372637663
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.392936448,
          "watts": {
            "core": 6.030256370328441,
            "dram": 1.9069508742412837,
            "package-0": 26.730312716784375,
            "psys": 43.516597670570256,
            "uncore": 16.29327395177468
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1116.0949352903115,
            "sha256": "0b8aeac4667d48fb5fafee7dd6a8d3dfdcbf32877fc72e9ba5308bdb34b0f5fd",
            "sum": 350539.43440529774,
            "top1": 42
          },
          "gpu_active_fraction": 0.9852941176470589,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 7.086042,
            "dram": 2.086664,
            "package-0": 17.477921,
            "psys": 28.96599,
            "uncore": 6.177597
          },
          "net_j": {
            "core": 4.500730314460494,
            "dram": 1.7481872617516214,
            "package-0": 13.552600092376515,
            "psys": 22.00438142288327,
            "uncore": 6.17625828466606
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.685439648,
          "token_latencies_s": [
            0.052803102,
            0.049185031,
            0.04858575,
            0.048914197,
            0.048747139,
            0.048780339,
            0.048633463,
            0.048553321,
            0.048760136,
            0.048563576,
            0.048401535,
            0.048464501,
            0.0485082,
            0.048509397
          ],
          "watts": {
            "core": 10.33795173751023,
            "dram": 3.0442709377675214,
            "package-0": 25.498847420043024,
            "psys": 42.2589940405665,
            "uncore": 9.012605293588152
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.48672,
            "dram": 11.041720000000002,
            "package-0": 141.423221,
            "psys": 232.011674,
            "uncore": 79.04050799999999
          },
          "net_j": {
            "core": 17.190983884726204,
            "dram": 8.384541525515937,
            "package-0": 110.60787376179333,
            "psys": 177.36024786470094,
            "uncore": 79.02999854641918
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.625,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.359491,
            "dram": 0.173034,
            "package-0": 3.068717,
            "psys": 4.300099,
            "uncore": 0.339477
          },
          "logits": {
            "l2": 1104.323434566963,
            "sha256": "4cc0c5b7c789d0230c86bfae04035aec63fca0279550b97617a206b017f84bae",
            "sum": 343667.50144172134,
            "top1": 261
          },
          "net_j": {
            "core": 2.071547591592009,
            "dram": 0.1353355877647577,
            "package-0": 2.6315278256508394,
            "psys": 3.5247382135581202,
            "uncore": 0.33932789834301275
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.076341986,
          "watts": {
            "core": 30.90685903822308,
            "dram": 2.266564037252057,
            "package-0": 40.19697627462822,
            "psys": 56.326789821789546,
            "uncore": 4.446792882752618
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1522.4001689327483,
          "decode_tokens_per_s": 19.82405508809927,
          "end_to_end_ms": 5380.976807,
          "handoff_estimate_ms": 28.140056499999996,
          "handoff_plus_first_decode_ms": 76.341986,
          "output_tokens_per_s_e2e": 2.973437829946774,
          "prefill_tokens_per_s": 1768.0436139441701,
          "steady_tpot_ms": 48.201929500000006,
          "steady_tpot_p95_ms": 50.6232435,
          "ttft_ms": 4624.3203140000005
        },
        "prefill": {
          "gpu_active_fraction": 0.8512035010940919,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 457,
          "gross_j": {
            "core": 29.064134,
            "dram": 8.792153,
            "package-0": 121.875054,
            "psys": 199.814491,
            "uncore": 72.484799
          },
          "logits": {
            "l2": 1063.3971159220778,
            "sha256": "084ff5e6e0b62105635ea609877c5eba14bc7d2005fb94d197b9e77e6987b4d0",
            "sum": 342411.9841043374,
            "top1": 220
          },
          "net_j": {
            "core": 11.622322193015183,
            "dram": 6.508618831724931,
            "package-0": 95.39286659654798,
            "psys": 152.84798117639912,
            "uncore": 72.47576735362316
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.624320314,
          "watts": {
            "core": 6.285060728169964,
            "dram": 1.9012854653216829,
            "package-0": 26.355236169740813,
            "psys": 43.20948321747247,
            "uncore": 15.674692512228077
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1116.0949352903115,
            "sha256": "0b8aeac4667d48fb5fafee7dd6a8d3dfdcbf32877fc72e9ba5308bdb34b0f5fd",
            "sum": 350539.43440529774,
            "top1": 42
          },
          "gpu_active_fraction": 0.9852941176470589,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 6.063095,
            "dram": 2.076533,
            "package-0": 16.47945,
            "psys": 27.897084,
            "uncore": 6.216232
          },
          "net_j": {
            "core": 3.4971141001190116,
            "dram": 1.7405871060262486,
            "package-0": 12.583479339594517,
            "psys": 20.98752847474369,
            "uncore": 6.214903294453014
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.680314507,
          "token_latencies_s": [
            0.053028419,
            0.049328149,
            0.04883711,
            0.048416965,
            0.048310765,
            0.048325123,
            0.04813078,
            0.048040704,
            0.048273079,
            0.047893402,
            0.04780936,
            0.048052681,
            0.047877043,
            0.047967316
          ],
          "watts": {
            "core": 8.912194194912265,
            "dram": 3.0523132736900465,
            "package-0": 24.223281776938503,
            "psys": 41.006157759325866,
            "uncore": 9.137291555653979
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.480738,
            "dram": 11.055086000000001,
            "package-0": 141.684817,
            "psys": 233.68049399999998,
            "uncore": 79.26127199999999
          },
          "net_j": {
            "core": 17.140400181094368,
            "dram": 8.392068137077759,
            "package-0": 110.80175026755944,
            "psys": 178.90896644548164,
            "uncore": 79.2507394509507
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.75,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 8,
          "gross_j": {
            "core": 2.132685,
            "dram": 0.173523,
            "package-0": 2.874199,
            "psys": 4.100027,
            "uncore": 0.368163
          },
          "logits": {
            "l2": 1104.323434566963,
            "sha256": "4cc0c5b7c789d0230c86bfae04035aec63fca0279550b97617a206b017f84bae",
            "sum": 343667.50144172134,
            "top1": 261
          },
          "net_j": {
            "core": 1.8378931070757456,
            "dram": 0.13492796368883872,
            "package-0": 2.4266116598870795,
            "psys": 3.3062249289674286,
            "uncore": 0.3680103520892719
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.078157714,
          "watts": {
            "core": 27.28694188778346,
            "dram": 2.220164729997093,
            "package-0": 36.77434833879609,
            "psys": 52.45837922025201,
            "uncore": 4.710513923168223
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1519.0618939371043,
          "decode_tokens_per_s": 19.745044508515853,
          "end_to_end_ms": 5392.801987000001,
          "handoff_estimate_ms": 29.980421500000006,
          "handoff_plus_first_decode_ms": 78.157714,
          "output_tokens_per_s_e2e": 2.966917761595907,
          "prefill_tokens_per_s": 1764.686443149489,
          "steady_tpot_ms": 48.17729249999999,
          "steady_tpot_p95_ms": 50.75145075,
          "ttft_ms": 4633.117703
        },
        "prefill": {
          "gpu_active_fraction": 0.8518518518518519,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 459,
          "gross_j": {
            "core": 29.451218,
            "dram": 8.804664,
            "package-0": 122.497367,
            "psys": 201.727816,
            "uncore": 72.677488
          },
          "logits": {
            "l2": 1063.3971159220778,
            "sha256": "084ff5e6e0b62105635ea609877c5eba14bc7d2005fb94d197b9e77e6987b4d0",
            "sum": 342411.9841043374,
            "top1": 220
          },
          "net_j": {
            "core": 11.976224580818837,
            "dram": 6.516785595813701,
            "package-0": 95.9647994134422,
            "psys": 154.67195625493287,
            "uncore": 72.66843917165887
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 4.633117703,
          "watts": {
            "core": 6.356673818351298,
            "dram": 1.9003756356759238,
            "package-0": 26.439511113797405,
            "psys": 43.54040387736724,
            "uncore": 15.686518810635965
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1116.0949352903115,
            "sha256": "0b8aeac4667d48fb5fafee7dd6a8d3dfdcbf32877fc72e9ba5308bdb34b0f5fd",
            "sum": 350539.43440529774,
            "top1": 42
          },
          "gpu_active_fraction": 0.9852941176470589,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 68,
          "gross_j": {
            "core": 5.896835,
            "dram": 2.076899,
            "package-0": 16.313251,
            "psys": 27.852651,
            "uncore": 6.215621
          },
          "net_j": {
            "core": 3.3262824931997836,
            "dram": 1.7403545775752192,
            "package-0": 12.410339194230167,
            "psys": 20.93078526158134,
            "uncore": 6.21428992720255
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.68152657,
          "token_latencies_s": [
            0.053352234,
            0.049351029,
            0.048413264,
            0.048992454,
            0.048325679,
            0.048244088,
            0.048131646,
            0.048064406,
            0.048222939,
            0.048081485,
            0.048072649,
            0.048090432,
            0.048070211,
            0.048094232
          ],
          "watts": {
            "core": 8.652391938292295,
            "dram": 3.047421907556737,
            "package-0": 23.936338975016042,
            "psys": 40.86803394913862,
            "uncore": 9.120144794941742
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.789744785055518,
    "compile_target": "AUTO:GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393809,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06067449878901243,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 14.149258,
            "dram": 5.502182,
            "package-0": 47.26111,
            "psys": 80.940162,
            "uncore": 21.727911000000002
          },
          "net_j": {
            "core": 9.25251221111787,
            "dram": 4.523871932324662,
            "package-0": 38.697736013214964,
            "psys": 61.14214097649398,
            "uncore": 21.723677138400266
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.629938,
            "dram": 0.29724,
            "package-0": 4.014516,
            "psys": 5.900986,
            "uncore": 0.761778
          },
          "logits": {
            "l2": 656.1393639359858,
            "sha256": "759589a4c665f287600ccbb81e67d5a11ba7b9e7dbbe102b502bb439588346fd",
            "sum": -851.9276055010268,
            "top1": 78
          },
          "net_j": {
            "core": 2.3513703327458884,
            "dram": 0.24158557864484417,
            "package-0": 3.5273599951407357,
            "psys": 4.7747097401035345,
            "uncore": 0.7615371427041213
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.113399307,
          "watts": {
            "core": 23.191834849572757,
            "dram": 2.6211800394864846,
            "package-0": 35.401592004437916,
            "psys": 52.03723158555104,
            "uncore": 6.7176600999863245
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 513.7037132624425,
          "decode_tokens_per_s": 10.62354782209812,
          "end_to_end_ms": 1993.366942,
          "handoff_estimate_ms": 22.278633500000005,
          "handoff_plus_first_decode_ms": 113.39930700000001,
          "output_tokens_per_s_e2e": 8.026620519725665,
          "prefill_tokens_per_s": 1733.7185288979988,
          "steady_tpot_ms": 91.1206735,
          "steady_tpot_p95_ms": 97.85417325,
          "ttft_ms": 581.409256
        },
        "prefill": {
          "gpu_active_fraction": 0.8793103448275862,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 2.795648,
            "dram": 1.294064,
            "package-0": 15.784567,
            "psys": 25.852656,
            "uncore": 10.118444
          },
          "logits": {
            "l2": 687.4985252735918,
            "sha256": "406295ad6f2a0cd6b351a0545dc583118a18409a1827387988bed8f0ee7f4e57",
            "sum": 121544.93127582863,
            "top1": 290
          },
          "net_j": {
            "core": 1.3674045261323107,
            "dram": 1.008718378679654,
            "package-0": 13.286870869095447,
            "psys": 20.078128274562786,
            "uncore": 10.117209101261166
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.581409256,
          "watts": {
            "core": 4.8083995415442775,
            "dram": 2.2257368396625594,
            "package-0": 27.14880583187688,
            "psys": 44.465504690898825,
            "uncore": 17.403307387318236
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 597.2912714442998,
            "sha256": "3ec855524c524afa8a2eb18bffab2133436dcb0e5f21c04bcd06f32483d1c0dd",
            "sum": -69328.94927807688,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 8.723672,
            "dram": 3.910878,
            "package-0": 27.462027,
            "psys": 49.18652,
            "uncore": 10.847689
          },
          "net_j": {
            "core": 5.533737352239671,
            "dram": 3.2735679750001636,
            "package-0": 21.883505148978784,
            "psys": 36.28930296182766,
            "uncore": 10.844930894434977
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.298558379,
          "token_latencies_s": [
            0.098024314,
            0.093376952,
            0.09067414,
            0.091202439,
            0.090443544,
            0.097762559,
            0.09307435,
            0.090874063,
            0.091028076,
            0.091038908,
            0.090533161,
            0.097218945,
            0.09289667,
            0.090351357
          ],
          "watts": {
            "core": 6.717966739945853,
            "dram": 3.011707492898169,
            "package-0": 21.14808809839423,
            "psys": 37.877788781339035,
            "uncore": 8.353639832776437
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.710773,
            "dram": 5.541489,
            "package-0": 50.114374,
            "psys": 85.053188,
            "uncore": 21.803411
          },
          "net_j": {
            "core": 11.759481709699099,
            "dram": 4.55228140660349,
            "package-0": 41.45561145352985,
            "psys": 65.03463419199078,
            "uncore": 21.79912997685383
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.620233,
            "dram": 0.294433,
            "package-0": 3.938162,
            "psys": 5.836472,
            "uncore": 0.702757
          },
          "logits": {
            "l2": 656.1393639359858,
            "sha256": "759589a4c665f287600ccbb81e67d5a11ba7b9e7dbbe102b502bb439588346fd",
            "sum": -851.9276055010268,
            "top1": 78
          },
          "net_j": {
            "core": 2.3485753835244947,
            "dram": 0.24015912248688048,
            "package-0": 3.463090214260142,
            "psys": 4.738133749648354,
            "uncore": 0.702522117324799
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.110586364,
          "watts": {
            "core": 23.69399720927618,
            "dram": 2.6624711162399732,
            "package-0": 35.6116419561457,
            "psys": 52.777501573340444,
            "uncore": 6.354825084944468
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 508.0445377105066,
          "decode_tokens_per_s": 10.824217659499535,
          "end_to_end_ms": 2015.571321,
          "handoff_estimate_ms": 20.241441499999997,
          "handoff_plus_first_decode_ms": 110.586364,
          "output_tokens_per_s_e2e": 7.938195901726665,
          "prefill_tokens_per_s": 1600.5338288425166,
          "steady_tpot_ms": 90.34492250000001,
          "steady_tpot_p95_ms": 95.07992044999999,
          "ttft_ms": 629.789875
        },
        "prefill": {
          "gpu_active_fraction": 0.8253968253968254,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 63,
          "gross_j": {
            "core": 3.693227,
            "dram": 1.357723,
            "package-0": 17.143022,
            "psys": 28.166676,
            "uncore": 10.389378
          },
          "logits": {
            "l2": 687.4985252735918,
            "sha256": "406295ad6f2a0cd6b351a0545dc583118a18409a1827387988bed8f0ee7f4e57",
            "sum": 121544.93127582863,
            "top1": 290
          },
          "net_j": {
            "core": 2.14613556725422,
            "dram": 1.048633006456557,
            "package-0": 14.437485894688388,
            "psys": 21.91163438025128,
            "uncore": 10.388040342034733
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.629789875,
          "watts": {
            "core": 5.864220983228731,
            "dram": 2.1558349123983613,
            "package-0": 27.220224840864578,
            "psys": 44.72392637306212,
            "uncore": 16.496578323047824
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 597.2912714442998,
            "sha256": "3ec855524c524afa8a2eb18bffab2133436dcb0e5f21c04bcd06f32483d1c0dd",
            "sum": -69328.94927807688,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 126,
          "gross_j": {
            "core": 10.397313,
            "dram": 3.889333,
            "package-0": 29.03319,
            "psys": 51.05004,
            "uncore": 10.711276
          },
          "net_j": {
            "core": 7.264770758920383,
            "dram": 3.2634892776600526,
            "package-0": 23.55503534458132,
            "psys": 38.38486606209114,
            "uncore": 10.708567517494297
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.275195082,
          "token_latencies_s": [
            0.100339075,
            0.092248068,
            0.090524258,
            0.090702288,
            0.090640303,
            0.0900161,
            0.08998771,
            0.089685271,
            0.090268508,
            0.090421337,
            0.089856105,
            0.089977806,
            0.090461312,
            0.090021764
          ],
          "watts": {
            "core": 8.153507762665603,
            "dram": 3.0499905895967063,
            "package-0": 22.767645836952813,
            "psys": 40.03312177140282,
            "uncore": 8.399715581713638
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.943255999999998,
            "dram": 5.523424,
            "package-0": 50.047846,
            "psys": 85.594995,
            "uncore": 21.611761
          },
          "net_j": {
            "core": 12.037326304696698,
            "dram": 4.543279099866579,
            "package-0": 41.46841130091035,
            "psys": 65.75984254671522,
            "uncore": 21.60751919774122
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.513971,
            "dram": 0.289795,
            "package-0": 3.798575,
            "psys": 5.668564,
            "uncore": 0.688597
          },
          "logits": {
            "l2": 656.1393639359858,
            "sha256": "759589a4c665f287600ccbb81e67d5a11ba7b9e7dbbe102b502bb439588346fd",
            "sum": -851.9276055010268,
            "top1": 78
          },
          "net_j": {
            "core": 2.2551898694181722,
            "dram": 0.23809368727361402,
            "package-0": 3.3460214680346123,
            "psys": 4.622286638537196,
            "uncore": 0.6883732506900002
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.105344605,
          "watts": {
            "core": 23.864259588803815,
            "dram": 2.750923979448212,
            "package-0": 36.05856227758413,
            "psys": 53.80972286146025,
            "uncore": 6.53661381140496
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 512.7420592796457,
          "decode_tokens_per_s": 10.791038897709017,
          "end_to_end_ms": 1997.1055259999998,
          "handoff_estimate_ms": 14.368366999999992,
          "handoff_plus_first_decode_ms": 105.34460499999999,
          "output_tokens_per_s_e2e": 8.011594676244464,
          "prefill_tokens_per_s": 1660.4529404286563,
          "steady_tpot_ms": 90.976238,
          "steady_tpot_p95_ms": 95.74199429999999,
          "ttft_ms": 607.063275
        },
        "prefill": {
          "gpu_active_fraction": 0.8666666666666667,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 60,
          "gross_j": {
            "core": 4.003286,
            "dram": 1.336849,
            "package-0": 17.168413,
            "psys": 28.243458,
            "uncore": 10.17429
          },
          "logits": {
            "l2": 687.4985252735918,
            "sha256": "406295ad6f2a0cd6b351a0545dc583118a18409a1827387988bed8f0ee7f4e57",
            "sum": 121544.93127582863,
            "top1": 290
          },
          "net_j": {
            "core": 2.5120229150113835,
            "dram": 1.0389128292252452,
            "package-0": 14.560508881628438,
            "psys": 22.214135837674092,
            "uncore": 10.173000612765527
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.607063275,
          "watts": {
            "core": 6.594511914758804,
            "dram": 2.202157592221338,
            "package-0": 28.281093103515445,
            "psys": 46.52473500394172,
            "uncore": 16.759850939755825
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 597.2912714442998,
            "sha256": "3ec855524c524afa8a2eb18bffab2133436dcb0e5f21c04bcd06f32483d1c0dd",
            "sum": -69328.94927807688,
            "top1": 42
          },
          "gpu_active_fraction": 0.9921259842519685,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 127,
          "gross_j": {
            "core": 10.425999,
            "dram": 3.89678,
            "package-0": 29.080858,
            "psys": 51.682973,
            "uncore": 10.748874
          },
          "net_j": {
            "core": 7.270113520267143,
            "dram": 3.26627258336772,
            "package-0": 23.561880951247296,
            "psys": 38.92342007050394,
            "uncore": 10.746145334285693
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.284697646,
          "token_latencies_s": [
            0.100463359,
            0.093199721,
            0.090998013,
            0.091349208,
            0.091189707,
            0.091013846,
            0.091266685,
            0.090549693,
            0.090732896,
            0.090680358,
            0.090588082,
            0.090954463,
            0.090857153,
            0.090827806
          ],
          "watts": {
            "core": 8.115527441388338,
            "dram": 3.0332273217226713,
            "package-0": 22.636344116100297,
            "psys": 40.22967829116735,
            "uncore": 8.36685116802962
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.953581932932138,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393877,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05905019957572222,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 34.644442999999995,
            "dram": 14.288355,
            "package-0": 105.003394,
            "psys": 211.935675,
            "uncore": 31.031598
          },
          "net_j": {
            "core": 19.863919423877668,
            "dram": 11.335386713824334,
            "package-0": 79.15538053150351,
            "psys": 152.1765759749107,
            "uncore": 31.022432305796265
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 32,
          "gross_j": {
            "core": 4.542591,
            "dram": 0.465209,
            "package-0": 5.770127,
            "psys": 9.328345,
            "uncore": 0.0
          },
          "logits": {
            "l2": 657.1396888861135,
            "sha256": "d92c9a2b6fabdb458067c8b3cf593700f9212fa4e571c6361c9a75d8d32712e2",
            "sum": -1004.8725463692099,
            "top1": 78
          },
          "net_j": {
            "core": 3.7491938886851486,
            "dram": 0.3066979367085064,
            "package-0": 4.3826430510374355,
            "psys": 6.120563063379139,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 128055,
          "npu_utilization": 0.3964845169537374,
          "seconds": 0.322976042,
          "watts": {
            "core": 14.064792459126116,
            "dram": 1.4403823798175097,
            "package-0": 17.865495422722404,
            "psys": 28.88246738747266,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 170.18859187875364,
          "decode_tokens_per_s": 8.815723876501492,
          "end_to_end_ms": 6016.85453,
          "handoff_estimate_ms": 225.1610995,
          "handoff_plus_first_decode_ms": 322.976042,
          "output_tokens_per_s_e2e": 2.6591967481055256,
          "prefill_tokens_per_s": 233.58479753314208,
          "steady_tpot_ms": 97.8149425,
          "steady_tpot_p95_ms": 101.43412945,
          "ttft_ms": 4315.349332000001
        },
        "prefill": {
          "gpu_active_fraction": 0.9976580796252927,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 427,
          "gross_j": {
            "core": 23.23578,
            "dram": 9.758581,
            "package-0": 79.169414,
            "psys": 164.841375,
            "uncore": 31.031598
          },
          "logits": {
            "l2": 687.8061052992338,
            "sha256": "d7e8d20b5dc6aa0d71807dad56aa5d9dc2be40a56a5857dab5cf9f63f533a459",
            "sum": 121705.75016708858,
            "top1": 290
          },
          "net_j": {
            "core": 12.63503799721308,
            "dram": 7.6406820782912614,
            "package-0": 60.63095555819298,
            "psys": 121.98154071731193,
            "uncore": 31.022432305796265
          },
          "npu_busy_delta_us": 11610,
          "npu_utilization": 0.0026903963287299403,
          "seconds": 4.315349332,
          "watts": {
            "core": 5.384449371849833,
            "dram": 2.2613652451346895,
            "package-0": 18.34600351191221,
            "psys": 38.19884841712277,
            "uncore": 7.190981682499857
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 597.4529443816389,
            "sha256": "6f77105b6ab09a679ccc06c0136f724e643c4d6fe6374032a2408f24c7dcdb18",
            "sum": -68897.5107829012,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 137,
          "gross_j": {
            "core": 6.866072,
            "dram": 4.064565,
            "package-0": 20.063853,
            "psys": 37.765955,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.4796875379794407,
            "dram": 3.3880066988245674,
            "package-0": 14.141781922273099,
            "psys": 24.07447219421962,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1364402,
          "npu_utilization": 0.9897520078276821,
          "seconds": 1.378529156,
          "token_latencies_s": [
            0.105533086,
            0.099226999,
            0.097978259,
            0.097780159,
            0.097753358,
            0.097863056,
            0.0977643,
            0.097956971,
            0.09779655,
            0.097586589,
            0.09765897,
            0.097833297,
            0.097944115,
            0.097796588
          ],
          "watts": {
            "core": 4.980723091793642,
            "dram": 2.9484795314695544,
            "package-0": 14.554536559979733,
            "psys": 27.395833331217553,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 29.673021000000002,
            "dram": 14.110742,
            "package-0": 99.67113099999999,
            "psys": 201.55435500000002,
            "uncore": 31.065655
          },
          "net_j": {
            "core": 15.38405315862766,
            "dram": 11.255980552582702,
            "package-0": 74.68274467440527,
            "psys": 143.7826637739639,
            "uncore": 31.056220958985488
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.07692307692307693,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.351557,
            "dram": 0.312499,
            "package-0": 3.151786,
            "psys": 4.862475,
            "uncore": 0.002502
          },
          "logits": {
            "l2": 657.1396888861135,
            "sha256": "d92c9a2b6fabdb458067c8b3cf593700f9212fa4e571c6361c9a75d8d32712e2",
            "sum": -1004.8725463692099,
            "top1": 78
          },
          "net_j": {
            "core": 2.0346546918578547,
            "dram": 0.24918578508308087,
            "package-0": 2.597590806110611,
            "psys": 3.5812080464977907,
            "uncore": 0.002227997515256498
          },
          "npu_busy_delta_us": 124545,
          "npu_utilization": 0.9654309145792276,
          "seconds": 0.12900457,
          "watts": {
            "core": 18.22847826243675,
            "dram": 2.4223870518695576,
            "package-0": 24.431584090393073,
            "psys": 37.69226935138809,
            "uncore": 0.01939466175500604
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 176.04326096722374,
          "decode_tokens_per_s": 9.972969920543374,
          "end_to_end_ms": 5816.752055,
          "handoff_estimate_ms": 31.265180500000017,
          "handoff_plus_first_decode_ms": 129.00457,
          "output_tokens_per_s_e2e": 2.750675952612871,
          "prefill_tokens_per_s": 233.72901957483646,
          "steady_tpot_ms": 97.7393895,
          "steady_tpot_p95_ms": 100.5588236,
          "ttft_ms": 4312.686554
        },
        "prefill": {
          "gpu_active_fraction": 0.9976580796252927,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 427,
          "gross_j": {
            "core": 20.750008,
            "dram": 9.745886,
            "package-0": 76.719591,
            "psys": 159.886432,
            "uncore": 31.063153
          },
          "logits": {
            "l2": 687.8061052992338,
            "sha256": "d7e8d20b5dc6aa0d71807dad56aa5d9dc2be40a56a5857dab5cf9f63f533a459",
            "sum": 121705.75016708858,
            "top1": 290
          },
          "net_j": {
            "core": 10.155807164663742,
            "dram": 7.629293923744799,
            "package-0": 58.192571678295415,
            "psys": 117.05304429556816,
            "uncore": 31.053992961470232
          },
          "npu_busy_delta_us": 11586,
          "npu_utilization": 0.0026864924809464836,
          "seconds": 4.312686554,
          "watts": {
            "core": 4.811387922628982,
            "dram": 2.259817836972346,
            "package-0": 17.789280542274252,
            "psys": 37.073510907419404,
            "uncore": 7.202738388485258
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 597.4529443816389,
            "sha256": "6f77105b6ab09a679ccc06c0136f724e643c4d6fe6374032a2408f24c7dcdb18",
            "sum": -68897.5107829012,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 136,
          "gross_j": {
            "core": 6.571456,
            "dram": 4.052357,
            "package-0": 19.799754,
            "psys": 36.805448,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.1935913021060633,
            "dram": 3.3775008437548224,
            "package-0": 13.89258218999925,
            "psys": 23.148411431897948,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1363223,
          "npu_utilization": 0.9913909771319073,
          "seconds": 1.375060931,
          "token_latencies_s": [
            0.104538979,
            0.098415663,
            0.097789404,
            0.097858491,
            0.097865387,
            0.097795309,
            0.097663261,
            0.097792864,
            0.097606031,
            0.097689375,
            0.097618236,
            0.097591444,
            0.097370825,
            0.097432287
          ],
          "watts": {
            "core": 4.779028951990492,
            "dram": 2.94703813383234,
            "package-0": 14.399183013367137,
            "psys": 26.76641243325384,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 34.1312,
            "dram": 14.117762,
            "package-0": 104.07700299999999,
            "psys": 204.64516700000001,
            "uncore": 31.006513
          },
          "net_j": {
            "core": 19.822727888387618,
            "dram": 11.259103837343407,
            "package-0": 79.0545078258749,
            "psys": 146.79461810625787,
            "uncore": 30.997352823388496
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 14,
          "gross_j": {
            "core": 2.519647,
            "dram": 0.316894,
            "package-0": 3.324942,
            "psys": 5.038073,
            "uncore": 0.0
          },
          "logits": {
            "l2": 657.1396888861135,
            "sha256": "d92c9a2b6fabdb458067c8b3cf593700f9212fa4e571c6361c9a75d8d32712e2",
            "sum": -1004.8725463692099,
            "top1": 78
          },
          "net_j": {
            "core": 2.185195971223005,
            "dram": 0.25007476479694,
            "package-0": 2.740057800745655,
            "psys": 3.685854856657956,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 125828,
          "npu_utilization": 0.9241980851145687,
          "seconds": 0.136148302,
          "watts": {
            "core": 18.506635506919505,
            "dram": 2.327564834411229,
            "package-0": 24.421472402938967,
            "psys": 37.00430285204732,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 175.80329157642794,
          "decode_tokens_per_s": 9.921026729013935,
          "end_to_end_ms": 5824.691852,
          "handoff_estimate_ms": 38.46611899999999,
          "handoff_plus_first_decode_ms": 136.148302,
          "output_tokens_per_s_e2e": 2.7469264308816865,
          "prefill_tokens_per_s": 233.7254963120047,
          "steady_tpot_ms": 97.68218300000001,
          "steady_tpot_p95_ms": 100.82995355,
          "ttft_ms": 4312.7515650000005
        },
        "prefill": {
          "gpu_active_fraction": 0.9976580796252927,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 427,
          "gross_j": {
            "core": 23.51526,
            "dram": 9.735571,
            "package-0": 79.446513,
            "psys": 162.074475,
            "uncore": 31.006513
          },
          "logits": {
            "l2": 687.8061052992338,
            "sha256": "d7e8d20b5dc6aa0d71807dad56aa5d9dc2be40a56a5857dab5cf9f63f533a459",
            "sum": 121705.75016708858,
            "top1": 290
          },
          "net_j": {
            "core": 12.92089946384109,
            "dram": 7.6189470174688045,
            "package-0": 60.91921439529211,
            "psys": 119.24044160955938,
            "uncore": 30.997352823388496
          },
          "npu_busy_delta_us": 9091,
          "npu_utilization": 0.002107935006917098,
          "seconds": 4.312751565,
          "watts": {
            "core": 5.452495847624833,
            "dram": 2.2573920276346824,
            "package-0": 18.421305239268982,
            "psys": 37.580294750875595,
            "uncore": 7.189496666497645
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 597.4529443816389,
            "sha256": "6f77105b6ab09a679ccc06c0136f724e643c4d6fe6374032a2408f24c7dcdb18",
            "sum": -68897.5107829012,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 136,
          "gross_j": {
            "core": 8.096293,
            "dram": 4.065297,
            "package-0": 21.305548,
            "psys": 37.532619,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.716632453323522,
            "dram": 3.390082055077663,
            "package-0": 15.395235629837128,
            "psys": 23.868321640040527,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1363433,
          "npu_utilization": 0.9910168214855533,
          "seconds": 1.375791985,
          "token_latencies_s": [
            0.105289027,
            0.098428914,
            0.097896288,
            0.0978713,
            0.097644656,
            0.097664091,
            0.097479362,
            0.0974379,
            0.097747341,
            0.097683766,
            0.097648207,
            0.0976806,
            0.097716086,
            0.097568683
          ],
          "watts": {
            "core": 5.8848234967730235,
            "dram": 2.954877659067043,
            "package-0": 15.48602421898831,
            "psys": 27.280736775043792,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 23.223629924468696,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394012,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU+GPU",
      "key": "npu_gpu_combined",
      "label": "NPU + iGPU (subgraph combination)",
      "mode": "combined",
      "prefill_device": "NPU+GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06119983829557896,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 378.001717,
            "dram": 144.44207300000002,
            "package-0": 1133.3849169999999,
            "psys": 2402.3283229999997,
            "uncore": 369.285493
          },
          "net_j": {
            "core": 212.50720218401017,
            "dram": 111.37828799463438,
            "package-0": 843.9699699654745,
            "psys": 1733.2178515025864,
            "uncore": 369.14240187347315
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.85,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 500,
          "gross_j": {
            "core": 34.278232,
            "dram": 9.749242,
            "package-0": 83.419404,
            "psys": 173.291915,
            "uncore": 23.588441
          },
          "logits": {
            "l2": 3028.7603501122676,
            "sha256": "f451a70dcddf4cbe6d2ff66a3945a7621931ab3de01bc712804d7d9e5f08ec13",
            "sum": 1122318.0844391808,
            "top1": 141473
          },
          "net_j": {
            "core": 21.835099529569668,
            "dram": 7.263255940262374,
            "package-0": 61.65899382037975,
            "psys": 122.98311628572388,
            "uncore": 23.57768232413273
          },
          "npu_busy_delta_us": 11630,
          "npu_utilization": 0.0022959916109781073,
          "seconds": 5.065349518,
          "watts": {
            "core": 6.767199751604585,
            "dram": 1.9246928499910085,
            "package-0": 16.46863729809059,
            "psys": 34.2112453216106,
            "uncore": 4.656823959763718
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 15.199757511286032,
          "decode_tokens_per_s": 0.22554565483834121,
          "end_to_end_ms": 67369.49581200001,
          "handoff_estimate_ms": 677.0697624999995,
          "handoff_plus_first_decode_ms": 5065.349518,
          "output_tokens_per_s_e2e": 0.23749621111384425,
          "prefill_tokens_per_s": 1166.5134038170588,
          "steady_tpot_ms": 4388.2797555,
          "steady_tpot_p95_ms": 4420.6474222,
          "ttft_ms": 864.113517
        },
        "prefill": {
          "gpu_active_fraction": 0.13953488372093023,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 86,
          "gross_j": {
            "core": 2.475396,
            "dram": 1.562984,
            "package-0": 18.398391,
            "psys": 31.476177,
            "uncore": 0.766905
          },
          "logits": {
            "l2": 690.102758738694,
            "sha256": "2f570dc830d3dd55b4b024060856ce9cff17aab7732962671d383c10d232e4d0",
            "sum": 123462.46132551879,
            "top1": 290
          },
          "net_j": {
            "core": 0.3526838508557719,
            "dram": 1.138892009968671,
            "package-0": 14.686215856931373,
            "psys": 22.893844657768486,
            "uncore": 0.7650696444714443
          },
          "npu_busy_delta_us": 708564,
          "npu_utilization": 0.8199894875617365,
          "seconds": 0.864113517,
          "watts": {
            "core": 2.864665291423742,
            "dram": 1.808771613047224,
            "package-0": 21.291636617229308,
            "psys": 36.42597457481966,
            "uncore": 0.8875049225737316
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 2808.0053142516367,
            "sha256": "c21cea26a3148945c34eeca07ac73bd0f71652d4375b55a5b70de31d97ad04c0",
            "sum": -711887.96540685,
            "top1": 220
          },
          "gpu_active_fraction": 0.9996698580389568,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6058,
          "gross_j": {
            "core": 341.248089,
            "dram": 133.129847,
            "package-0": 1031.567122,
            "psys": 2197.560231,
            "uncore": 344.930147
          },
          "net_j": {
            "core": 190.31941880358474,
            "dram": 102.97614004440334,
            "package-0": 767.6247602881634,
            "psys": 1587.340890559094,
            "uncore": 344.799649904869
          },
          "npu_busy_delta_us": 163479,
          "npu_utilization": 0.0026607895961474515,
          "seconds": 61.440032777,
          "token_latencies_s": [
            4.357933267,
            4.358327048,
            4.376426719,
            4.393156215,
            4.383403296,
            4.368133214,
            4.365546084,
            4.375708867,
            4.419513096,
            4.422754028,
            4.419285121,
            4.400771137,
            4.400375828,
            4.398633877
          ],
          "watts": {
            "core": 5.554165152199362,
            "dram": 2.166825780891136,
            "package-0": 16.789820502605036,
            "psys": 35.76756280349274,
            "uncore": 5.614094449655375
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 372.833519,
            "dram": 145.401666,
            "package-0": 1132.461397,
            "psys": 2418.4415010000002,
            "uncore": 372.374596
          },
          "net_j": {
            "core": 208.22301999969795,
            "dram": 112.51449656949946,
            "package-0": 844.5924068859165,
            "psys": 1752.9051917030906,
            "uncore": 372.23226921795356
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9977011494252873,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 435,
          "gross_j": {
            "core": 27.147025,
            "dram": 9.56217,
            "package-0": 76.55852,
            "psys": 160.024553,
            "uncore": 24.625303
          },
          "logits": {
            "l2": 2808.823057621367,
            "sha256": "0bac2cc2cf3bdce12f64c9fb8052874bc0f3588cdfb36a3d50ac586531a65ff8",
            "sum": -712076.476965785,
            "top1": 220
          },
          "net_j": {
            "core": 16.332964979774964,
            "dram": 7.401652740915023,
            "package-0": 57.64701340740075,
            "psys": 116.30225315834392,
            "uncore": 24.615952865149048
          },
          "npu_busy_delta_us": 11686,
          "npu_utilization": 0.0026545897501502125,
          "seconds": 4.402186816,
          "watts": {
            "core": 6.166713530042065,
            "dram": 2.1721408926231267,
            "package-0": 17.391020236066236,
            "psys": 36.3511499372043,
            "uncore": 5.593879594227561
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 15.281385512635882,
          "decode_tokens_per_s": 0.22676615859774746,
          "end_to_end_ms": 67009.630714,
          "handoff_estimate_ms": 0.0,
          "handoff_plus_first_decode_ms": 4402.186816,
          "output_tokens_per_s_e2e": 0.23877164863493566,
          "prefill_tokens_per_s": 1169.1096864390988,
          "steady_tpot_ms": 4408.0777025,
          "steady_tpot_p95_ms": 4430.79879175,
          "ttft_ms": 862.19455
        },
        "prefill": {
          "gpu_active_fraction": 0.13953488372093023,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 86,
          "gross_j": {
            "core": 3.31591,
            "dram": 1.56158,
            "package-0": 19.191724,
            "psys": 31.622051,
            "uncore": 0.779905
          },
          "logits": {
            "l2": 690.102758738694,
            "sha256": "2f570dc830d3dd55b4b024060856ce9cff17aab7732962671d383c10d232e4d0",
            "sum": 123462.46132551879,
            "top1": 290
          },
          "net_j": {
            "core": 1.1979118316980681,
            "dram": 1.1384298058297748,
            "package-0": 15.487792613634195,
            "psys": 23.058777742279688,
            "uncore": 0.7780737203094253
          },
          "npu_busy_delta_us": 712074,
          "npu_utilization": 0.8258855266482489,
          "seconds": 0.86219455,
          "watts": {
            "core": 3.845895337659001,
            "dram": 1.8111689525293333,
            "package-0": 22.259157170501716,
            "psys": 36.676236239257136,
            "uncore": 0.904558025795918
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 2806.3459274564684,
            "sha256": "1263a5ad985e73d4ff10f0150c722a4dc401619b5ca1690100a3cadbbd76a777",
            "sum": -709671.8661493696,
            "top1": 220
          },
          "gpu_active_fraction": 0.9998359310910583,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6095,
          "gross_j": {
            "core": 342.370584,
            "dram": 134.277916,
            "package-0": 1036.711153,
            "psys": 2226.794897,
            "uncore": 346.969388
          },
          "net_j": {
            "core": 190.69214318822492,
            "dram": 103.97441402275466,
            "package-0": 771.4576008648816,
            "psys": 1613.544160802467,
            "uncore": 346.8382426324951
          },
          "npu_busy_delta_us": 162473,
          "npu_utilization": 0.00263134413927608,
          "seconds": 61.745249348,
          "token_latencies_s": [
            4.400253701,
            4.402764825,
            4.405152867,
            4.404330115,
            4.400407505,
            4.401590521,
            4.411365182,
            4.418415966,
            4.409351803,
            4.406803602,
            4.410526606,
            4.412189366,
            4.43028982,
            4.431744025
          ],
          "watts": {
            "core": 5.544889487292836,
            "dram": 2.174708458025677,
            "package-0": 16.790136309224902,
            "psys": 36.06423037422115,
            "uncore": 5.619369776036684
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 355.299016,
            "dram": 145.67620100000002,
            "package-0": 1115.5413330000001,
            "psys": 2417.44126,
            "uncore": 372.594749
          },
          "net_j": {
            "core": 190.51473839882365,
            "dram": 112.75431272516659,
            "package-0": 827.3684408223124,
            "psys": 1751.2023468633415,
            "uncore": 372.45227196417756
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9977011494252873,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 435,
          "gross_j": {
            "core": 24.970883,
            "dram": 9.614966,
            "package-0": 74.543083,
            "psys": 159.563923,
            "uncore": 24.720762
          },
          "logits": {
            "l2": 2805.755314989439,
            "sha256": "59e00002720744e4e9ec04d1081d48ec008f2032aa2805aeb701fc6b48dd571c",
            "sum": -709422.8614426963,
            "top1": 220
          },
          "net_j": {
            "core": 14.132508103934388,
            "dram": 7.449590925366763,
            "package-0": 55.589054825876566,
            "psys": 115.743315742801,
            "uncore": 24.711390841836106
          },
          "npu_busy_delta_us": 9454,
          "npu_utilization": 0.0021427511447029237,
          "seconds": 4.412084914,
          "watts": {
            "core": 5.6596560326309255,
            "dram": 2.1792341234165105,
            "package-0": 16.895205884063362,
            "psys": 36.165197658296925,
            "uncore": 5.602966053884972
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 15.265270032248647,
          "decode_tokens_per_s": 0.22651740861171837,
          "end_to_end_ms": 67080.372495,
          "handoff_estimate_ms": 0.6553485000004855,
          "handoff_plus_first_decode_ms": 4412.084914,
          "output_tokens_per_s_e2e": 0.23851984425388512,
          "prefill_tokens_per_s": 1171.6889881071302,
          "steady_tpot_ms": 4411.4295655,
          "steady_tpot_p95_ms": 4431.34374395,
          "ttft_ms": 860.296555
        },
        "prefill": {
          "gpu_active_fraction": 0.13953488372093023,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 86,
          "gross_j": {
            "core": 2.698784,
            "dram": 1.561397,
            "package-0": 18.556043,
            "psys": 31.281902,
            "uncore": 0.779844
          },
          "logits": {
            "l2": 690.102758738694,
            "sha256": "2f570dc830d3dd55b4b024060856ce9cff17aab7732962671d383c10d232e4d0",
            "sum": 123462.46132551879,
            "top1": 290
          },
          "net_j": {
            "core": 0.5854482944026236,
            "dram": 1.1391783089954863,
            "package-0": 14.8602652759969,
            "psys": 22.737479533933133,
            "uncore": 0.7780167516034023
          },
          "npu_busy_delta_us": 711986,
          "npu_utilization": 0.8276053133794079,
          "seconds": 0.860296555,
          "watts": {
            "core": 3.137039180634636,
            "dram": 1.8149520545272901,
            "package-0": 21.56935639478412,
            "psys": 36.36176597266509,
            "uncore": 0.9064827651204531
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": null,
            "sha256": "b0ba3595a28b4c7ae5c9e581e843cac4c1d56ce6f3684ebf9bc0b9924c710c7a",
            "sum": null,
            "top1": 0
          },
          "gpu_active_fraction": 0.9995077932731747,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6095,
          "gross_j": {
            "core": 327.629349,
            "dram": 134.499838,
            "package-0": 1022.442207,
            "psys": 2226.595435,
            "uncore": 347.094143
          },
          "net_j": {
            "core": 175.79678200048664,
            "dram": 104.16554349080434,
            "package-0": 756.919120720439,
            "psys": 1612.7215515866073,
            "uncore": 346.96286437073803
          },
          "npu_busy_delta_us": 164109,
          "npu_utilization": 0.0026551421147302185,
          "seconds": 61.807991026,
          "token_latencies_s": [
            4.429692753,
            4.429075659,
            4.43440987,
            4.417047683,
            4.410237764,
            4.411556353,
            4.408600612,
            4.405753037,
            4.413138022,
            4.405813664,
            4.41057777,
            4.41405597,
            4.406664255,
            4.411302778
          ],
          "watts": {
            "core": 5.300760363852955,
            "dram": 2.1760914044823365,
            "package-0": 16.54223329423378,
            "psys": 36.02439422539014,
            "uncore": 5.615683946983363
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.413142912089825,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 393724,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06183239724487066,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 15.497032,
            "dram": 5.491015,
            "package-0": 48.586788999999996,
            "psys": 82.980745,
            "uncore": 21.675908999999997
          },
          "net_j": {
            "core": 10.580342034643227,
            "dram": 4.508720329186607,
            "package-0": 39.9885368616069,
            "psys": 63.102087724957705,
            "uncore": 21.671657894114656
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.616327,
            "dram": 0.298584,
            "package-0": 4.050954,
            "psys": 6.01268,
            "uncore": 0.810423
          },
          "logits": {
            "l2": 656.1393639359858,
            "sha256": "759589a4c665f287600ccbb81e67d5a11ba7b9e7dbbe102b502bb439588346fd",
            "sum": -851.9276055010268,
            "top1": 78
          },
          "net_j": {
            "core": 2.3313956586923443,
            "dram": 0.24165819420660734,
            "package-0": 3.5526692734957956,
            "psys": 4.860674784970748,
            "uncore": 0.8101766404958446
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.115989831,
          "watts": {
            "core": 22.55652049359396,
            "dram": 2.5742256663862197,
            "package-0": 34.92507890627067,
            "psys": 51.837992590919455,
            "uncore": 6.98701768088618
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 511.61991347332133,
          "decode_tokens_per_s": 10.56431005256959,
          "end_to_end_ms": 2001.485816,
          "handoff_estimate_ms": 24.994537999999995,
          "handoff_plus_first_decode_ms": 115.989831,
          "output_tokens_per_s_e2e": 7.994061148020646,
          "prefill_tokens_per_s": 1733.1177749760836,
          "steady_tpot_ms": 90.995293,
          "steady_tpot_p95_ms": 100.76107875000001,
          "ttft_ms": 581.6107910000001
        },
        "prefill": {
          "gpu_active_fraction": 0.8793103448275862,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 58,
          "gross_j": {
            "core": 2.7445,
            "dram": 1.290097,
            "package-0": 15.724325,
            "psys": 26.085871,
            "uncore": 10.109715
          },
          "logits": {
            "l2": 687.4985252735918,
            "sha256": "406295ad6f2a0cd6b351a0545dc583118a18409a1827387988bed8f0ee7f4e57",
            "sum": 121544.93127582863,
            "top1": 290
          },
          "net_j": {
            "core": 1.3157614513712272,
            "dram": 1.0046524687823668,
            "package-0": 13.225763087918855,
            "psys": 20.309341638996898,
            "uncore": 10.108479673205858
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.581610791,
          "watts": {
            "core": 4.718791402204228,
            "dram": 2.218144883078691,
            "package-0": 27.035820592262706,
            "psys": 44.85107808118368,
            "uncore": 17.382268617502316
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 597.2912714442998,
            "sha256": "3ec855524c524afa8a2eb18bffab2133436dcb0e5f21c04bcd06f32483d1c0dd",
            "sum": -69328.94927807688,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 10.136205,
            "dram": 3.902334,
            "package-0": 28.81151,
            "psys": 50.882194,
            "uncore": 10.755771
          },
          "net_j": {
            "core": 6.933184924579656,
            "dram": 3.262409666197633,
            "package-0": 23.21010450019225,
            "psys": 37.93207130099006,
            "uncore": 10.753001580412954
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.303885194,
          "token_latencies_s": [
            0.107377861,
            0.093277404,
            0.090525707,
            0.097188518,
            0.097198196,
            0.093032097,
            0.090834573,
            0.091046992,
            0.090177748,
            0.090048414,
            0.091091117,
            0.09023503,
            0.090943594,
            0.090855458
          ],
          "watts": {
            "core": 7.773847764084666,
            "dram": 2.992850918130757,
            "package-0": 22.09666167894226,
            "psys": 39.023523109351295,
            "uncore": 8.249016899259306
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.58236,
            "dram": 5.532701,
            "package-0": 47.632873000000004,
            "psys": 81.486486,
            "uncore": 21.544745
          },
          "net_j": {
            "core": 9.619657569928716,
            "dram": 4.541213600094776,
            "package-0": 38.9541547805825,
            "psys": 61.42179584045928,
            "uncore": 21.54045411046734
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.490106,
            "dram": 0.290161,
            "package-0": 3.78594,
            "psys": 5.6046,
            "uncore": 0.701781
          },
          "logits": {
            "l2": 656.1393639359858,
            "sha256": "759589a4c665f287600ccbb81e67d5a11ba7b9e7dbbe102b502bb439588346fd",
            "sum": -851.9276055010268,
            "top1": 78
          },
          "net_j": {
            "core": 2.2310315437328234,
            "dram": 0.23840108438029334,
            "package-0": 3.332873503373618,
            "psys": 4.557136694170133,
            "uncore": 0.7015569970725144
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.105464012,
          "watts": {
            "core": 23.610954606961094,
            "dram": 2.7512797446014097,
            "package-0": 35.8979326521354,
            "psys": 53.1422984363614,
            "uncore": 6.6542224849174145
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 506.87635015320603,
          "decode_tokens_per_s": 10.790380945119539,
          "end_to_end_ms": 2020.2165670000004,
          "handoff_estimate_ms": 15.022169000000002,
          "handoff_plus_first_decode_ms": 105.464012,
          "output_tokens_per_s_e2e": 7.919942971143844,
          "prefill_tokens_per_s": 1599.7725859785992,
          "steady_tpot_ms": 90.44184299999999,
          "steady_tpot_p95_ms": 97.91285529999999,
          "ttft_ms": 630.089557
        },
        "prefill": {
          "gpu_active_fraction": 0.8387096774193549,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 62,
          "gross_j": {
            "core": 3.291557,
            "dram": 1.35046,
            "package-0": 16.502094,
            "psys": 27.198722,
            "uncore": 10.167637
          },
          "logits": {
            "l2": 687.4985252735918,
            "sha256": "406295ad6f2a0cd6b351a0545dc583118a18409a1827387988bed8f0ee7f4e57",
            "sum": 121544.93127582863,
            "top1": 290
          },
          "net_j": {
            "core": 1.7437293924232928,
            "dram": 1.0412229277067675,
            "package-0": 13.79527048042008,
            "psys": 20.940703953577717,
            "uncore": 10.166298705517649
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.630089557,
          "watts": {
            "core": 5.223951045422579,
            "dram": 2.143282625457003,
            "package-0": 26.190076976628895,
            "psys": 43.166438322671645,
            "uncore": 16.136812437283417
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 597.2912714442998,
            "sha256": "3ec855524c524afa8a2eb18bffab2133436dcb0e5f21c04bcd06f32483d1c0dd",
            "sum": -69328.94927807688,
            "top1": 42
          },
          "gpu_active_fraction": 0.9921259842519685,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 127,
          "gross_j": {
            "core": 8.800697,
            "dram": 3.89208,
            "package-0": 27.344839,
            "psys": 48.683164,
            "uncore": 10.675327
          },
          "net_j": {
            "core": 5.6448966337726,
            "dram": 3.2615895880077144,
            "package-0": 21.8260107967888,
            "psys": 35.92395519271143,
            "uncore": 10.672598407877178
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.284662998,
          "token_latencies_s": [
            0.107271905,
            0.092873367,
            0.09061633,
            0.090533028,
            0.090398671,
            0.090062548,
            0.090588847,
            0.09034045,
            0.090045882,
            0.090937211,
            0.090102587,
            0.089988632,
            0.090485015,
            0.090387272
          ],
          "watts": {
            "core": 6.850588063718793,
            "dram": 3.0296505823389492,
            "package-0": 21.285612680190233,
            "psys": 37.895669195572175,
            "uncore": 8.309826792411437
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.736291,
            "dram": 5.52678,
            "package-0": 48.065308,
            "psys": 81.769688,
            "uncore": 21.879095
          },
          "net_j": {
            "core": 9.78143912396,
            "dram": 4.536861045012433,
            "package-0": 39.40031874121712,
            "psys": 61.73673839642411,
            "uncore": 21.874810898273097
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.490839,
            "dram": 0.290893,
            "package-0": 3.784048,
            "psys": 5.581651,
            "uncore": 0.690367
          },
          "logits": {
            "l2": 656.1393639359858,
            "sha256": "759589a4c665f287600ccbb81e67d5a11ba7b9e7dbbe102b502bb439588346fd",
            "sum": -851.9276055010268,
            "top1": 78
          },
          "net_j": {
            "core": 2.229550008135565,
            "dram": 0.23869064718558702,
            "package-0": 3.327108748416759,
            "psys": 4.525234110667522,
            "uncore": 0.6901410823238975
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.106365505,
          "watts": {
            "core": 23.41773303290385,
            "dram": 2.7348434062340043,
            "package-0": 35.575894647423524,
            "psys": 52.47613876322028,
            "uncore": 6.490515886705939
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 507.67945391364344,
          "decode_tokens_per_s": 10.60815137197216,
          "end_to_end_ms": 2017.020764,
          "handoff_estimate_ms": 15.092098999999997,
          "handoff_plus_first_decode_ms": 106.365505,
          "output_tokens_per_s_e2e": 7.932491467400679,
          "prefill_tokens_per_s": 1671.6035576181314,
          "steady_tpot_ms": 91.273406,
          "steady_tpot_p95_ms": 100.62771444999998,
          "ttft_ms": 603.0137920000001
        },
        "prefill": {
          "gpu_active_fraction": 0.8666666666666667,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 60,
          "gross_j": {
            "core": 3.405082,
            "dram": 1.330563,
            "package-0": 16.748798,
            "psys": 27.170401,
            "uncore": 10.388218
          },
          "logits": {
            "l2": 687.4985252735918,
            "sha256": "406295ad6f2a0cd6b351a0545dc583118a18409a1827387988bed8f0ee7f4e57",
            "sum": 121544.93127582863,
            "top1": 290
          },
          "net_j": {
            "core": 1.9237665510484818,
            "dram": 1.034614245542988,
            "package-0": 14.158290195699612,
            "psys": 21.181298100810807,
            "uncore": 10.386937213766283
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.603013792,
          "watts": {
            "core": 5.646773001172086,
            "dram": 2.2065216710665214,
            "package-0": 27.77514912959072,
            "psys": 45.05767755308654,
            "uncore": 17.227164847334038
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 597.2912714442998,
            "sha256": "3ec855524c524afa8a2eb18bffab2133436dcb0e5f21c04bcd06f32483d1c0dd",
            "sum": -69328.94927807688,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 8.84037,
            "dram": 3.905324,
            "package-0": 27.532462,
            "psys": 49.017636,
            "uncore": 10.80051
          },
          "net_j": {
            "core": 5.628122564775953,
            "dram": 3.2635561522838574,
            "package-0": 21.91491979710075,
            "psys": 36.03020618494578,
            "uncore": 10.797732602182917
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.307641467,
          "token_latencies_s": [
            0.106311982,
            0.093507582,
            0.091153165,
            0.091108806,
            0.091988969,
            0.091393647,
            0.090797245,
            0.090891614,
            0.097223945,
            0.093223965,
            0.090617826,
            0.090980076,
            0.090845977,
            0.097566955
          ],
          "watts": {
            "core": 6.760545778868298,
            "dram": 2.9865403465367466,
            "package-0": 21.05505422917274,
            "psys": 37.48553195736183,
            "uncore": 8.25953464505726
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.800072302110493,
    "compile_target": "AUTO:GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 398915,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05963654909282923,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 19.056958,
            "dram": 6.960127,
            "package-0": 68.072764,
            "psys": 115.775948,
            "uncore": 34.181492
          },
          "net_j": {
            "core": 8.613791821207752,
            "dram": 5.59287644953272,
            "package-0": 52.21673442620212,
            "psys": 87.65506971318558,
            "uncore": 34.17608436316588
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.592889,
            "dram": 0.305725,
            "package-0": 4.024465,
            "psys": 6.088607,
            "uncore": 0.788206
          },
          "logits": {
            "l2": 597.2273383364098,
            "sha256": "fcc036c7b26185e79a5e482f5d15937fe9f09db4f7711fbfd4b32d0318d977c9",
            "sum": 14783.873326078989,
            "top1": 220
          },
          "net_j": {
            "core": 2.124192088020137,
            "dram": 0.2443617984636539,
            "package-0": 3.312834832456317,
            "psys": 4.826521480193037,
            "uncore": 0.7879633013162989
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.12426488,
          "watts": {
            "core": 20.865823070846726,
            "dram": 2.4602687420613134,
            "package-0": 32.386181839953494,
            "psys": 48.997005429048016,
            "uncore": 6.3429506389898735
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 739.6758694038298,
          "decode_tokens_per_s": 9.826272362787046,
          "end_to_end_ms": 2768.7803329999997,
          "handoff_estimate_ms": 25.534078999999988,
          "handoff_plus_first_decode_ms": 124.26487999999999,
          "output_tokens_per_s_e2e": 5.77871772971742,
          "prefill_tokens_per_s": 1635.727819476029,
          "steady_tpot_ms": 98.73080100000001,
          "steady_tpot_p95_ms": 103.9432763,
          "ttft_ms": 1242.260464
        },
        "prefill": {
          "gpu_active_fraction": 0.8943089430894309,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 123,
          "gross_j": {
            "core": 6.861677,
            "dram": 2.652581,
            "package-0": 34.933871,
            "psys": 57.282263,
            "uncore": 22.105229
          },
          "logits": {
            "l2": 597.5347857334565,
            "sha256": "f657289f689ae92500218d2a246d30ac2917dc3da7ea94d5cbc4c64bf5fc7387",
            "sum": -5077.78165544162,
            "top1": 220
          },
          "net_j": {
            "core": 2.1761725891683,
            "dram": 2.03914074871527,
            "package-0": 27.819793216082186,
            "psys": 44.66535190297019,
            "uncore": 22.10280277161249
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.242260464,
          "watts": {
            "core": 5.52354131749942,
            "dram": 2.135285696414146,
            "package-0": 28.12121291175536,
            "psys": 46.111314543131115,
            "uncore": 17.79435926731707
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 683.1319075602754,
            "sha256": "8dff8948fa4761bb522f9e112d6d62b7d86c314faa7abb52288c95eeaa37d5b0",
            "sum": -135169.30241355416,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 9.602392,
            "dram": 4.001821,
            "package-0": 29.114428,
            "psys": 52.405078,
            "uncore": 11.288057
          },
          "net_j": {
            "core": 4.3134271440193155,
            "dram": 3.309373902353796,
            "package-0": 21.084106377663623,
            "psys": 38.16319633002236,
            "uncore": 11.285318290237084
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.402254989,
          "token_latencies_s": [
            0.105260344,
            0.10150523,
            0.09848226,
            0.102534748,
            0.100741341,
            0.098661505,
            0.098800097,
            0.103234086,
            0.101156855,
            0.09852638,
            0.098390305,
            0.09821515,
            0.098378143,
            0.098339222
          ],
          "watts": {
            "core": 6.847821598301334,
            "dram": 2.853846861941883,
            "package-0": 20.76257758281365,
            "psys": 37.37200324555237,
            "uncore": 8.049931780274807
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 22.018987,
            "dram": 7.075055000000001,
            "package-0": 71.567077,
            "psys": 120.049009,
            "uncore": 34.424839
          },
          "net_j": {
            "core": 11.349929749193517,
            "dram": 5.678230114769733,
            "package-0": 55.36807332649007,
            "psys": 91.3198616084489,
            "uncore": 34.419314393191385
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.479303,
            "dram": 0.297789,
            "package-0": 3.84844,
            "psys": 5.751145,
            "uncore": 0.74243
          },
          "logits": {
            "l2": 597.2273383364098,
            "sha256": "fcc036c7b26185e79a5e482f5d15937fe9f09db4f7711fbfd4b32d0318d977c9",
            "sum": 14783.873326078989,
            "top1": 220
          },
          "net_j": {
            "core": 2.049909865060237,
            "dram": 0.24157156650693667,
            "package-0": 3.1964853983182655,
            "psys": 4.5948948859918355,
            "uncore": 0.742207653434498
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.113844331,
          "watts": {
            "core": 21.77801018480226,
            "dram": 2.6157560713321772,
            "package-0": 33.8044061236567,
            "psys": 50.517623051428004,
            "uncore": 6.5214490126873335
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 724.0150503497318,
          "decode_tokens_per_s": 9.859794128458285,
          "end_to_end_ms": 2828.6704799999998,
          "handoff_estimate_ms": 13.979001000000006,
          "handoff_plus_first_decode_ms": 113.84433100000001,
          "output_tokens_per_s_e2e": 5.65636758085728,
          "prefill_tokens_per_s": 1554.3004567468697,
          "steady_tpot_ms": 99.86533,
          "steady_tpot_p95_ms": 104.1105031,
          "ttft_ms": 1307.340541
        },
        "prefill": {
          "gpu_active_fraction": 0.8682170542635659,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 8.67051,
            "dram": 2.768304,
            "package-0": 37.264614,
            "psys": 60.559842,
            "uncore": 22.334965
          },
          "logits": {
            "l2": 597.5347857334565,
            "sha256": "f657289f689ae92500218d2a246d30ac2917dc3da7ea94d5cbc4c64bf5fc7387",
            "sum": -5077.78165544162,
            "top1": 220
          },
          "net_j": {
            "core": 3.739539363141189,
            "dram": 2.122726576241825,
            "package-0": 29.77784084069321,
            "psys": 47.28195072249102,
            "uncore": 22.332411665313245
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.307340541,
          "watts": {
            "core": 6.632174041943062,
            "dram": 2.1175079584715486,
            "package-0": 28.504137086956597,
            "psys": 46.32292818952671,
            "uncore": 17.084274754392396
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 683.1319075602754,
            "sha256": "8dff8948fa4761bb522f9e112d6d62b7d86c314faa7abb52288c95eeaa37d5b0",
            "sum": -135169.30241355416,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 10.869174,
            "dram": 4.008962,
            "package-0": 30.454023,
            "psys": 53.738022,
            "uncore": 11.347444
          },
          "net_j": {
            "core": 5.560480520992091,
            "dram": 3.3139319720209715,
            "package-0": 22.393747087478594,
            "psys": 39.44301599996605,
            "uncore": 11.344695074443642
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.407485608,
          "token_latencies_s": [
            0.106032921,
            0.101961515,
            0.100012575,
            0.099967872,
            0.099074879,
            0.099419853,
            0.103075355,
            0.101230266,
            0.099911529,
            0.098849471,
            0.099791116,
            0.098938158,
            0.099372649,
            0.099819131
          ],
          "watts": {
            "core": 7.722405073430775,
            "dram": 2.8483147374392197,
            "package-0": 21.63718252385853,
            "psys": 38.18015736328581,
            "uncore": 8.062209613727006
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.703191,
            "dram": 7.0545469999999995,
            "package-0": 71.128358,
            "psys": 119.914489,
            "uncore": 34.367832
          },
          "net_j": {
            "core": 11.079000426287372,
            "dram": 5.663596194326064,
            "package-0": 54.99747613425398,
            "psys": 91.30615653603785,
            "uncore": 34.36233062586733
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 11,
          "gross_j": {
            "core": 2.316766,
            "dram": 0.295531,
            "package-0": 3.672964,
            "psys": 5.557054,
            "uncore": 0.739806
          },
          "logits": {
            "l2": 597.2273383364098,
            "sha256": "fcc036c7b26185e79a5e482f5d15937fe9f09db4f7711fbfd4b32d0318d977c9",
            "sum": 14783.873326078989,
            "top1": 220
          },
          "net_j": {
            "core": 1.8915888707439317,
            "dram": 0.23986553858064646,
            "package-0": 3.0274106285256055,
            "psys": 4.4121565524780255,
            "uncore": 0.7395858365490091
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.112726548,
          "watts": {
            "core": 20.552088581653365,
            "dram": 2.6216628224967913,
            "package-0": 32.58295463815676,
            "psys": 49.29676370467762,
            "uncore": 6.562837353983376
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 727.0726149942805,
          "decode_tokens_per_s": 9.858521320098015,
          "end_to_end_ms": 2816.7750480000004,
          "handoff_estimate_ms": 12.829017999999984,
          "handoff_plus_first_decode_ms": 112.726548,
          "output_tokens_per_s_e2e": 5.680254804642816,
          "prefill_tokens_per_s": 1568.8106920415082,
          "steady_tpot_ms": 99.89753000000002,
          "steady_tpot_p95_ms": 104.039077,
          "ttft_ms": 1295.2486940000001
        },
        "prefill": {
          "gpu_active_fraction": 0.875,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 128,
          "gross_j": {
            "core": 8.4954,
            "dram": 2.75042,
            "package-0": 36.951627,
            "psys": 60.003326,
            "uncore": 22.244877
          },
          "logits": {
            "l2": 597.5347857334565,
            "sha256": "f657289f689ae92500218d2a246d30ac2917dc3da7ea94d5cbc4c64bf5fc7387",
            "sum": -5077.78165544162,
            "top1": 220
          },
          "net_j": {
            "core": 3.6100368698561427,
            "dram": 2.110813647427889,
            "package-0": 29.534100462156353,
            "psys": 46.848244522061265,
            "uncore": 22.2423472816025
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.295248694,
          "watts": {
            "core": 6.558894858843224,
            "dram": 2.1234686533488216,
            "package-0": 28.528596223390593,
            "psys": 46.32571820219106,
            "uncore": 17.174213031864287
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 683.1319075602754,
            "sha256": "8dff8948fa4761bb522f9e112d6d62b7d86c314faa7abb52288c95eeaa37d5b0",
            "sum": -135169.30241355416,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928571428571429,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 10.891025,
            "dram": 4.008596,
            "package-0": 30.503767,
            "psys": 54.354109,
            "uncore": 11.383149
          },
          "net_j": {
            "core": 5.577374685687298,
            "dram": 3.3129170083175286,
            "package-0": 22.43596504357202,
            "psys": 40.04575546149855,
            "uncore": 11.38039750771582
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.408799806,
          "token_latencies_s": [
            0.106193723,
            0.102547991,
            0.100181598,
            0.099243188,
            0.099948852,
            0.099824431,
            0.100020096,
            0.102878883,
            0.100572106,
            0.099846208,
            0.099497948,
            0.099507828,
            0.099387479,
            0.099125849
          ],
          "watts": {
            "core": 7.730711598351825,
            "dram": 2.8453978932475805,
            "package-0": 21.65230778005942,
            "psys": 38.58185440437234,
            "uncore": 8.080033054746176
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 6.534896406345069,
    "compile_target": "NPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 398983,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 2048,
      "NPUW_LLM_MIN_RESPONSE_LEN": 2048,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.059766506776213646,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 68.32496,
            "dram": 25.172114999999998,
            "package-0": 200.704382,
            "psys": 396.91188700000004,
            "uncore": 65.266922
          },
          "net_j": {
            "core": 27.041020580213505,
            "dram": 19.76709817552495,
            "package-0": 138.02230329360532,
            "psys": 285.7443908835781,
            "uncore": 65.24898366649069
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 35,
          "gross_j": {
            "core": 4.564686,
            "dram": 0.488768,
            "package-0": 5.849412,
            "psys": 9.550818,
            "uncore": 0.0
          },
          "logits": {
            "l2": 924.1045417971341,
            "sha256": "eabdef562e197dcc4559d67137a6e297994f30355c212d87eb2e14c7047a635e",
            "sum": -80822.31826195121,
            "top1": 4710
          },
          "net_j": {
            "core": 3.238641773551792,
            "dram": 0.3151583238237132,
            "package-0": 3.8360574185680343,
            "psys": 5.9801070052243155,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 133882,
          "npu_utilization": 0.380809583961391,
          "seconds": 0.351572034,
          "watts": {
            "core": 12.983643630767288,
            "dram": 1.3902357205124,
            "package-0": 16.637876265209425,
            "psys": 27.16603448612184,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 187.10806505361293,
          "decode_tokens_per_s": 8.518436657183571,
          "end_to_end_ms": 10945.546358,
          "handoff_estimate_ms": 251.839064,
          "handoff_plus_first_decode_ms": 351.572034,
          "output_tokens_per_s_e2e": 1.461781758231351,
          "prefill_tokens_per_s": 221.23845940293506,
          "steady_tpot_ms": 99.73297000000001,
          "steady_tpot_p95_ms": 104.46886035,
          "ttft_ms": 9184.659871
        },
        "prefill": {
          "gpu_active_fraction": 0.9988974641675854,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 907,
          "gross_j": {
            "core": 55.644572,
            "dram": 20.534615,
            "package-0": 173.266402,
            "psys": 349.057517,
            "uncore": 65.266922
          },
          "logits": {
            "l2": 2607.309964597262,
            "sha256": "4f901afd538c27dea9094e9b2e2d3c4f348a01d843d780b0f16c0dfd34e7033b",
            "sum": -638918.1009616219,
            "top1": 220
          },
          "net_j": {
            "core": 21.002268244320447,
            "dram": 15.999140981057852,
            "package-0": 120.66842707070805,
            "psys": 255.7743123846559,
            "uncore": 65.24898366649069
          },
          "npu_busy_delta_us": 11765,
          "npu_utilization": 0.0012809401943285092,
          "seconds": 9.184659871,
          "watts": {
            "core": 6.058424893413236,
            "dram": 2.2357512731458664,
            "package-0": 18.864759766126785,
            "psys": 38.00440320083357,
            "uncore": 7.106079366757641
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 852.6998990461121,
            "sha256": "e7b82e5d25e5c8a63909eb7510b07668bc1ad20828d519affcb9a9d9bc2c3791",
            "sum": -96127.76050041988,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 8.115702,
            "dram": 4.148732,
            "package-0": 21.588568,
            "psys": 38.303552,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.800110562341267,
            "dram": 3.4527988706433836,
            "package-0": 13.517818804329234,
            "psys": 23.9899714936979,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1394879,
          "npu_utilization": 0.9897571099414533,
          "seconds": 1.409314453,
          "token_latencies_s": [
            0.109713659,
            0.101644738,
            0.099920969,
            0.101382686,
            0.099727762,
            0.099479697,
            0.099478436,
            0.099566454,
            0.099518227,
            0.099624318,
            0.099738178,
            0.099830598,
            0.09971947,
            0.09991612
          ],
          "watts": {
            "core": 5.758616881224875,
            "dram": 2.943794403845513,
            "package-0": 15.318489038443147,
            "psys": 27.178854171589204,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 68.854865,
            "dram": 25.014585,
            "package-0": 200.975681,
            "psys": 396.23214,
            "uncore": 65.414323
          },
          "net_j": {
            "core": 28.509491123068294,
            "dram": 19.732447983539863,
            "package-0": 139.71864166860587,
            "psys": 287.5919700740218,
            "uncore": 65.39617692235339
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 16,
          "gross_j": {
            "core": 2.625482,
            "dram": 0.341369,
            "package-0": 3.504813,
            "psys": 5.45775,
            "uncore": 0.002625
          },
          "logits": {
            "l2": 918.1321595628507,
            "sha256": "cab5c133a903f9459bc22557fda52b8ad3eb3890e6cf5e2d76ac0ba74501ae02",
            "sum": -58615.71318984404,
            "top1": 1718
          },
          "net_j": {
            "core": 2.026030294952455,
            "dram": 0.2628869898822991,
            "package-0": 2.594655697074013,
            "psys": 3.843573978297077,
            "uncore": 0.002314594413705833
          },
          "npu_busy_delta_us": 122835,
          "npu_utilization": 0.7728791967699029,
          "seconds": 0.158931694,
          "watts": {
            "core": 16.519562171155112,
            "dram": 2.147897574161639,
            "package-0": 22.052322678949107,
            "psys": 34.34022417202701,
            "uncore": 0.016516529421752718
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 191.4608114969905,
          "decode_tokens_per_s": 9.587053232222363,
          "end_to_end_ms": 10696.705941999999,
          "handoff_estimate_ms": 58.99234650000002,
          "handoff_plus_first_decode_ms": 158.93169400000002,
          "output_tokens_per_s_e2e": 1.4957875898202382,
          "prefill_tokens_per_s": 222.5118995479228,
          "steady_tpot_ms": 99.9393475,
          "steady_tpot_p95_ms": 102.39185755,
          "ttft_ms": 9132.095874999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9988913525498891,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 902,
          "gross_j": {
            "core": 59.00052,
            "dram": 20.526498,
            "package-0": 176.60996,
            "psys": 352.033766,
            "uncore": 65.411698
          },
          "logits": {
            "l2": 2695.370592197745,
            "sha256": "07df53b12f154266b8efa14380ec825d8f02f7b9b1d08813636109b65b67466c",
            "sum": -667790.8857600205,
            "top1": 220
          },
          "net_j": {
            "core": 24.556474856138387,
            "dram": 16.016980591486348,
            "package-0": 124.31300435532205,
            "psys": 259.28442306958885,
            "uncore": 65.39386232793969
          },
          "npu_busy_delta_us": 11270,
          "npu_utilization": 0.0012341088129454183,
          "seconds": 9.132095875,
          "watts": {
            "core": 6.460786308816541,
            "dram": 2.2477313292552354,
            "package-0": 19.339477204076115,
            "psys": 38.54906593389221,
            "uncore": 7.162835223737728
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 836.8015178215198,
            "sha256": "568ae868543f3c53e0d210409c186a2477fc8751f54b903ed56f8bc232e53616",
            "sum": -19761.276002621278,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 7.228863,
            "dram": 4.146718,
            "package-0": 20.860908,
            "psys": 38.740624,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.9269859719774542,
            "dram": 3.4525804021712156,
            "package-0": 12.810981616209808,
            "psys": 24.463973026135857,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1388243,
          "npu_utilization": 0.9875964706188163,
          "seconds": 1.405678373,
          "token_latencies_s": [
            0.105429823,
            0.10075603,
            0.099763821,
            0.099695589,
            0.100276127,
            0.099792842,
            0.100388974,
            0.099979174,
            0.099883471,
            0.100276706,
            0.099996648,
            0.099744264,
            0.099899521,
            0.099754323
          ],
          "watts": {
            "core": 5.142615223262029,
            "dram": 2.949976381261434,
            "package-0": 14.840455968230222,
            "psys": 27.560091087778297,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 65.757034,
            "dram": 25.076473,
            "package-0": 198.52274699999998,
            "psys": 399.781997,
            "uncore": 65.88197199999999
          },
          "net_j": {
            "core": 25.25754212515117,
            "dram": 19.77415839430289,
            "package-0": 137.03170779816514,
            "psys": 290.7268252070558,
            "uncore": 65.86374292190342
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 17,
          "gross_j": {
            "core": 2.725274,
            "dram": 0.349913,
            "package-0": 3.628409,
            "psys": 5.779099,
            "uncore": 0.005371
          },
          "logits": {
            "l2": 911.9156589065695,
            "sha256": "108204db436c26fbbb8f3b21395d80af2310480d661184bea4fff12957b38ce5",
            "sum": -35362.86651572399,
            "top1": 1718
          },
          "net_j": {
            "core": 2.0690658106841706,
            "dram": 0.2640002612096108,
            "package-0": 2.632077401112863,
            "psys": 4.012091723984843,
            "uncore": 0.005031205007308373
          },
          "npu_busy_delta_us": 131007,
          "npu_utilization": 0.753002716665644,
          "seconds": 0.173979452,
          "watts": {
            "core": 15.664344085875154,
            "dram": 2.011231763162468,
            "package-0": 20.855388140893787,
            "psys": 33.21713532009516,
            "uncore": 0.030871461763197182
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 190.732220702847,
          "decode_tokens_per_s": 9.505573428216362,
          "end_to_end_ms": 10737.567006,
          "handoff_estimate_ms": 74.2844115,
          "handoff_plus_first_decode_ms": 173.979452,
          "output_tokens_per_s_e2e": 1.4900954742409922,
          "prefill_tokens_per_s": 221.8450704149606,
          "steady_tpot_ms": 99.6950405,
          "steady_tpot_p95_ms": 102.94850089999998,
          "ttft_ms": 9159.545425999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9988950276243094,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 905,
          "gross_j": {
            "core": 55.694987,
            "dram": 20.581612,
            "package-0": 173.92906,
            "psys": 354.108469,
            "uncore": 65.876601
          },
          "logits": {
            "l2": 2722.790925199642,
            "sha256": "0df34f45f9f926a49cf433f896f8897822b464feb4bddff4979e0fc2074631a3",
            "sum": -676477.0488090478,
            "top1": 220
          },
          "net_j": {
            "core": 21.147408824207446,
            "dram": 16.058539737290253,
            "package-0": 121.47490846395573,
            "psys": 261.0803370759303,
            "uncore": 65.85871171689611
          },
          "npu_busy_delta_us": 9406,
          "npu_utilization": 0.0010269068564582279,
          "seconds": 9.159545426,
          "watts": {
            "core": 6.0805405082555675,
            "dram": 2.247012383559743,
            "package-0": 18.988830985683023,
            "psys": 38.66004834637741,
            "uncore": 7.1921255844208964
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 858.7947249227907,
            "sha256": "5980de6d458600a7a9248f6eb7f4c5fa6366e9300635034843014fdbea2145dd",
            "sum": 37917.90754732303,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 7.336773,
            "dram": 4.144948,
            "package-0": 20.965278,
            "psys": 39.894429,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.041067490259551,
            "dram": 3.451618395803028,
            "package-0": 12.924721933096551,
            "psys": 25.63439640714067,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1389452,
          "npu_utilization": 0.9896084827448995,
          "seconds": 1.404042128,
          "token_latencies_s": [
            0.107559285,
            0.100465771,
            0.099748129,
            0.099710932,
            0.099729064,
            0.099807425,
            0.099494863,
            0.099565665,
            0.099489828,
            0.099675299,
            0.099679149,
            0.09965846,
            0.09985043,
            0.099574389
          ],
          "watts": {
            "core": 5.225465001147032,
            "dram": 2.952153583813263,
            "package-0": 14.932086140366867,
            "psys": 28.413982888695777,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 7.723630679771304,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 398828,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06377757806330919,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 20.016244,
            "dram": 6.974348,
            "package-0": 69.264837,
            "psys": 116.913763,
            "uncore": 34.382968
          },
          "net_j": {
            "core": 9.507867702090076,
            "dram": 5.598559945701675,
            "package-0": 53.30979783854081,
            "psys": 88.61728989918167,
            "uncore": 34.37752659633347
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.293024,
            "dram": 0.300475,
            "package-0": 3.704519,
            "psys": 5.687364,
            "uncore": 0.777647
          },
          "logits": {
            "l2": 597.2273383364098,
            "sha256": "fcc036c7b26185e79a5e482f5d15937fe9f09db4f7711fbfd4b32d0318d977c9",
            "sum": 14783.873326078989,
            "top1": 220
          },
          "net_j": {
            "core": 1.8385880440822373,
            "dram": 0.24097888548422558,
            "package-0": 3.014541474691437,
            "psys": 4.463679727746709,
            "uncore": 0.7774116858651964
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.120483895,
          "watts": {
            "core": 19.03178843944247,
            "dram": 2.4939017783248127,
            "package-0": 30.747005647518286,
            "psys": 47.204350423764105,
            "uncore": 6.4543647099058346
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 735.0857833443984,
          "decode_tokens_per_s": 9.772237343680946,
          "end_to_end_ms": 2786.0693899999997,
          "handoff_estimate_ms": 20.43445149999999,
          "handoff_plus_first_decode_ms": 120.48389499999999,
          "output_tokens_per_s_e2e": 5.742857682378112,
          "prefill_tokens_per_s": 1624.1594121798773,
          "steady_tpot_ms": 100.0494435,
          "steady_tpot_p95_ms": 105.15026845,
          "ttft_ms": 1251.108718
        },
        "prefill": {
          "gpu_active_fraction": 0.8870967741935484,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 124,
          "gross_j": {
            "core": 7.416301,
            "dram": 2.667535,
            "package-0": 35.762909,
            "psys": 58.124607,
            "uncore": 22.332523
          },
          "logits": {
            "l2": 597.5347857334565,
            "sha256": "f657289f689ae92500218d2a246d30ac2917dc3da7ea94d5cbc4c64bf5fc7387",
            "sum": -5077.78165544162,
            "top1": 220
          },
          "net_j": {
            "core": 2.69742312655993,
            "dram": 2.0497253951742955,
            "package-0": 28.598159743366086,
            "psys": 45.417829375525145,
            "uncore": 22.330079490305017
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.251108718,
          "watts": {
            "core": 5.927783008222951,
            "dram": 2.1321368491974653,
            "package-0": 28.584973060670496,
            "psys": 46.458478119245264,
            "uncore": 17.850185742211412
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 683.1319075602754,
            "sha256": "8dff8948fa4761bb522f9e112d6d62b7d86c314faa7abb52288c95eeaa37d5b0",
            "sum": -135169.30241355416,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928571428571429,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 10.306919,
            "dram": 4.006338,
            "package-0": 29.797409,
            "psys": 53.101792,
            "uncore": 11.272798
          },
          "net_j": {
            "core": 4.971856531447909,
            "dram": 3.3078556650431543,
            "package-0": 21.697096620483286,
            "psys": 38.73578079590982,
            "uncore": 11.270035420163255
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.414476777,
          "token_latencies_s": [
            0.109264903,
            0.10290229,
            0.101456903,
            0.099853759,
            0.102934696,
            0.101404243,
            0.100399433,
            0.099736826,
            0.099262246,
            0.099457353,
            0.09852163,
            0.099424135,
            0.100245128,
            0.099584195
          ],
          "watts": {
            "core": 7.2867361045405135,
            "dram": 2.832381602260834,
            "package-0": 21.066029138490407,
            "psys": 37.5416499326521,
            "uncore": 7.969588602160557
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.576423,
            "dram": 7.061261000000001,
            "package-0": 71.315247,
            "psys": 119.056703,
            "uncore": 34.658418999999995
          },
          "net_j": {
            "core": 10.893718929440029,
            "dram": 5.662649432311498,
            "package-0": 55.095523137431016,
            "psys": 90.29080807894302,
            "uncore": 34.652887326651985
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.466547,
            "dram": 0.298094,
            "package-0": 3.806386,
            "psys": 5.717454,
            "uncore": 0.719236
          },
          "logits": {
            "l2": 597.2273383364098,
            "sha256": "fcc036c7b26185e79a5e482f5d15937fe9f09db4f7711fbfd4b32d0318d977c9",
            "sum": 14783.873326078989,
            "top1": 220
          },
          "net_j": {
            "core": 2.0352238872556896,
            "dram": 0.24162388804041796,
            "package-0": 3.1515010815329374,
            "psys": 4.556006930393431,
            "uncore": 0.719012654061426
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.114356023,
          "watts": {
            "core": 21.569016963802596,
            "dram": 2.606718843309198,
            "package-0": 33.285400280140905,
            "psys": 49.99696430506332,
            "uncore": 6.289445725128094
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 723.0901438068121,
          "decode_tokens_per_s": 9.79450211911987,
          "end_to_end_ms": 2832.2886399999998,
          "handoff_estimate_ms": 14.458865500000002,
          "handoff_plus_first_decode_ms": 114.35602300000001,
          "output_tokens_per_s_e2e": 5.649141748490719,
          "prefill_tokens_per_s": 1562.094933725843,
          "steady_tpot_ms": 99.8971575,
          "steady_tpot_p95_ms": 106.0088787,
          "ttft_ms": 1300.817227
        },
        "prefill": {
          "gpu_active_fraction": 0.8682170542635659,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 8.781655,
            "dram": 2.758233,
            "package-0": 37.653712,
            "psys": 60.378264,
            "uncore": 22.656131
          },
          "logits": {
            "l2": 597.5347857334565,
            "sha256": "f657289f689ae92500218d2a246d30ac2917dc3da7ea94d5cbc4c64bf5fc7387",
            "sum": -5077.78165544162,
            "top1": 220
          },
          "net_j": {
            "core": 3.875288717385728,
            "dram": 2.1158768518883866,
            "package-0": 30.204296033125694,
            "psys": 47.16662619765273,
            "uncore": 22.653590405837523
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.300817227,
          "watts": {
            "core": 6.75087538643121,
            "dram": 2.120384741798934,
            "package-0": 28.94619722006495,
            "psys": 46.41563991218576,
            "uncore": 17.41684421896113
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 683.1319075602754,
            "sha256": "8dff8948fa4761bb522f9e112d6d62b7d86c314faa7abb52288c95eeaa37d5b0",
            "sum": -135169.30241355416,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928571428571429,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 10.328221,
            "dram": 4.004934,
            "package-0": 29.855149,
            "psys": 52.960985,
            "uncore": 11.283052
          },
          "net_j": {
            "core": 4.983206324798611,
            "dram": 3.3051486923826934,
            "package-0": 21.739726022772388,
            "psys": 38.568174950896854,
            "uncore": 11.28028426675304
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.41711539,
          "token_latencies_s": [
            0.11166375,
            0.102615279,
            0.099494242,
            0.100003437,
            0.100070121,
            0.099772248,
            0.09946119,
            0.099695025,
            0.100263253,
            0.099171416,
            0.102963948,
            0.102512974,
            0.099790878,
            0.099609278
          ],
          "watts": {
            "core": 7.288200433699333,
            "dram": 2.826117074347771,
            "package-0": 21.067549763890433,
            "psys": 37.372387156136945,
            "uncore": 7.9619853680369665
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.425297999999998,
            "dram": 7.067181,
            "package-0": 71.02679599999999,
            "psys": 119.97704900000001,
            "uncore": 34.505893
          },
          "net_j": {
            "core": 10.739367010413103,
            "dram": 5.668146954421139,
            "package-0": 54.802172653630116,
            "psys": 91.20246477958266,
            "uncore": 34.50035965570221
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.453424,
            "dram": 0.299011,
            "package-0": 3.805166,
            "psys": 5.726304,
            "uncore": 0.725218
          },
          "logits": {
            "l2": 597.2273383364098,
            "sha256": "fcc036c7b26185e79a5e482f5d15937fe9f09db4f7711fbfd4b32d0318d977c9",
            "sum": 14783.873326078989,
            "top1": 220
          },
          "net_j": {
            "core": 2.0189440550144955,
            "dram": 0.24212758612166932,
            "package-0": 3.1454880116421657,
            "psys": 4.556356357510966,
            "uncore": 0.7249930194036973
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.115192989,
          "watts": {
            "core": 21.298379539400614,
            "dram": 2.595739572310256,
            "package-0": 33.0329652267292,
            "psys": 49.71052535150381,
            "uncore": 6.295678289934816
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 722.8717862911677,
          "decode_tokens_per_s": 9.807851600405176,
          "end_to_end_ms": 2833.144188,
          "handoff_estimate_ms": 15.073428,
          "handoff_plus_first_decode_ms": 115.192989,
          "output_tokens_per_s_e2e": 5.647435830399748,
          "prefill_tokens_per_s": 1558.5723337212964,
          "steady_tpot_ms": 100.11956099999999,
          "steady_tpot_p95_ms": 104.67509199999999,
          "ttft_ms": 1303.757263
        },
        "prefill": {
          "gpu_active_fraction": 0.8682170542635659,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 8.331643,
            "dram": 2.757561,
            "package-0": 37.016507,
            "psys": 59.999358,
            "uncore": 22.451481
          },
          "logits": {
            "l2": 597.5347857334565,
            "sha256": "f657289f689ae92500218d2a246d30ac2917dc3da7ea94d5cbc4c64bf5fc7387",
            "sum": -5077.78165544162,
            "top1": 220
          },
          "net_j": {
            "core": 3.4141876164723417,
            "dram": 2.1137530336305477,
            "package-0": 29.55025427030691,
            "psys": 46.75785997597698,
            "uncore": 22.448934663725456
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.303757263,
          "watts": {
            "core": 6.3904863554344,
            "dram": 2.1150877377700943,
            "package-0": 28.39217701830743,
            "psys": 46.02034420267694,
            "uncore": 17.220598985073497
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 683.1319075602754,
            "sha256": "8dff8948fa4761bb522f9e112d6d62b7d86c314faa7abb52288c95eeaa37d5b0",
            "sum": -135169.30241355416,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928571428571429,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 10.640231,
            "dram": 4.010609,
            "package-0": 30.205123,
            "psys": 54.251387,
            "uncore": 11.329194
          },
          "net_j": {
            "core": 5.306235338926266,
            "dram": 3.3122663346689216,
            "package-0": 22.106430371681036,
            "psys": 39.888248446094714,
            "uncore": 11.326431972573062
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.414193936,
          "token_latencies_s": [
            0.106336986,
            0.102244007,
            0.100121094,
            0.100007831,
            0.103780226,
            0.101792269,
            0.099423881,
            0.101088198,
            0.099824522,
            0.100225698,
            0.099312594,
            0.099770934,
            0.100118028,
            0.100114329
          ],
          "watts": {
            "core": 7.523883909512111,
            "dram": 2.835968177988284,
            "package-0": 21.35854371249404,
            "psys": 38.36205602284523,
            "uncore": 8.01106107981501
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.763699364848435,
    "compile_target": "AUTO:GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 399323,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06079309061169624,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 31.347149,
            "dram": 10.57413,
            "package-0": 121.269038,
            "psys": 200.139563,
            "uncore": 66.86823700000001
          },
          "net_j": {
            "core": 14.18038655644472,
            "dram": 8.32660611702695,
            "package-0": 95.20446253205941,
            "psys": 153.91369347067837,
            "uncore": 66.85934777853947
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.160577,
            "dram": 0.309143,
            "package-0": 3.626272,
            "psys": 5.606187,
            "uncore": 0.795775
          },
          "logits": {
            "l2": 670.568131117269,
            "sha256": "f155fefe53ba4dfd4365752c28829b31e15f6eec90450cbf0ac8b52f5d1cfc31",
            "sum": -55835.01430409169,
            "top1": 1718
          },
          "net_j": {
            "core": 1.692335878432797,
            "dram": 0.24783947190769742,
            "package-0": 2.9153338664753936,
            "psys": 4.345328811684205,
            "uncore": 0.795532537331793
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.124144037,
          "watts": {
            "core": 17.40379201620453,
            "dram": 2.490196125972607,
            "package-0": 29.21019879513021,
            "psys": 45.15872961340866,
            "uncore": 6.410094429263647
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 899.9434864931993,
          "decode_tokens_per_s": 10.48851653020552,
          "end_to_end_ms": 4551.396906,
          "handoff_estimate_ms": 31.45649399999999,
          "handoff_plus_first_decode_ms": 124.144037,
          "output_tokens_per_s_e2e": 3.51540424411406,
          "prefill_tokens_per_s": 1307.1638322479619,
          "steady_tpot_ms": 92.687543,
          "steady_tpot_p95_ms": 96.28903589999999,
          "ttft_ms": 3121.26139
        },
        "prefill": {
          "gpu_active_fraction": 0.9158576051779935,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 309,
          "gross_j": {
            "core": 20.009715,
            "dram": 6.305403,
            "package-0": 89.136979,
            "psys": 143.999875,
            "uncore": 54.78294
          },
          "logits": {
            "l2": 662.3364776849841,
            "sha256": "d5055a398340419d05f270e2706e72a6086d083d704644e6d8b567ea276b1166",
            "sum": 8132.857491835253,
            "top1": 36
          },
          "net_j": {
            "core": 8.237075981035225,
            "dram": 4.764093888993046,
            "package-0": 71.26238912461984,
            "psys": 112.29905330483116,
            "uncore": 54.77684394107459
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.12126139,
          "watts": {
            "core": 6.410778368036648,
            "dram": 2.0201457718989695,
            "package-0": 28.557998790354436,
            "psys": 46.1351540314283,
            "uncore": 17.551538674561314
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 706.6311889009895,
            "sha256": "b5b9231e8728eabecddc2a8f58f01d40f25e30ef87c52a6cee95dc6ef8f92e49",
            "sum": -130420.66216149065,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 9.176857,
            "dram": 3.959584,
            "package-0": 28.505787,
            "psys": 50.533501,
            "uncore": 11.289522
          },
          "net_j": {
            "core": 4.250974696976697,
            "dram": 3.314672756126206,
            "package-0": 21.026739540964172,
            "psys": 37.26931135416299,
            "uncore": 11.286971300133086
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.305991479,
          "token_latencies_s": [
            0.101734693,
            0.093356759,
            0.092773563,
            0.092729729,
            0.092555365,
            0.09269699,
            0.092534318,
            0.092817279,
            0.092678096,
            0.09214686,
            0.092513595,
            0.092289964,
            0.092254252,
            0.0928761
          ],
          "watts": {
            "core": 7.026735738755919,
            "dram": 3.0318605164498167,
            "package-0": 21.82693184325133,
            "psys": 38.69359166010302,
            "uncore": 8.64440708958102
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 30.602338,
            "dram": 10.823580999999999,
            "package-0": 121.397212,
            "psys": 203.488066,
            "uncore": 67.27296
          },
          "net_j": {
            "core": 12.984179572425761,
            "dram": 8.51695900472438,
            "package-0": 94.64727464567737,
            "psys": 156.04669809514434,
            "uncore": 67.26383703855025
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.451471,
            "dram": 0.301208,
            "package-0": 3.807119,
            "psys": 5.721238,
            "uncore": 0.725706
          },
          "logits": {
            "l2": 670.568131117269,
            "sha256": "f155fefe53ba4dfd4365752c28829b31e15f6eec90450cbf0ac8b52f5d1cfc31",
            "sum": -55835.01430409169,
            "top1": 1718
          },
          "net_j": {
            "core": 2.0180797235955388,
            "dram": 0.2444671178683852,
            "package-0": 3.1490939549087535,
            "psys": 4.554221874269205,
            "uncore": 0.7254815831335296
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.114904352,
          "watts": {
            "core": 21.334883817107293,
            "dram": 2.621380258947894,
            "package-0": 33.13293999517094,
            "psys": 49.79130816559498,
            "uncore": 6.3157398946908465
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 876.8859758392337,
          "decode_tokens_per_s": 10.48239379444412,
          "end_to_end_ms": 4671.074817999999,
          "handoff_estimate_ms": 22.046770000000006,
          "handoff_plus_first_decode_ms": 114.904352,
          "output_tokens_per_s_e2e": 3.4253358431220065,
          "prefill_tokens_per_s": 1259.2188538799617,
          "steady_tpot_ms": 92.857582,
          "steady_tpot_p95_ms": 100.4056106,
          "ttft_ms": 3240.103964
        },
        "prefill": {
          "gpu_active_fraction": 0.897196261682243,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 321,
          "gross_j": {
            "core": 19.836985,
            "dram": 6.541548,
            "package-0": 89.900161,
            "psys": 147.508778,
            "uncore": 55.218059
          },
          "logits": {
            "l2": 662.3364776849841,
            "sha256": "d5055a398340419d05f270e2706e72a6086d083d704644e6d8b567ea276b1166",
            "sum": 8132.857491835253,
            "top1": 36
          },
          "net_j": {
            "core": 7.616100688837314,
            "dram": 4.941553274687056,
            "package-0": 71.3449929850285,
            "psys": 114.60094196753823,
            "uncore": 55.21173083256349
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.240103964,
          "watts": {
            "core": 6.122329783366173,
            "dram": 2.0189315135197927,
            "package-0": 27.746072965206867,
            "psys": 45.52593979666512,
            "uncore": 17.04206396261179
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 706.6311889009895,
            "sha256": "b5b9231e8728eabecddc2a8f58f01d40f25e30ef87c52a6cee95dc6ef8f92e49",
            "sum": -130420.66216149065,
            "top1": 42
          },
          "gpu_active_fraction": 0.9923076923076923,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 130,
          "gross_j": {
            "core": 8.313882,
            "dram": 3.980825,
            "package-0": 27.689932,
            "psys": 50.25805,
            "uncore": 11.329195
          },
          "net_j": {
            "core": 3.349999159992908,
            "dram": 3.33093861216894,
            "package-0": 20.153187705740116,
            "psys": 36.89153425333689,
            "uncore": 11.326624622853227
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.316066502,
          "token_latencies_s": [
            0.101705296,
            0.09275046,
            0.092622302,
            0.09970578,
            0.093185036,
            0.092897502,
            0.09279048,
            0.092689397,
            0.092686889,
            0.093199939,
            0.092817662,
            0.093313978,
            0.09301514,
            0.092660661
          ],
          "watts": {
            "core": 6.317220282839476,
            "dram": 3.024790156082857,
            "package-0": 21.03991854356916,
            "psys": 38.188077824049046,
            "uncore": 8.608375779478658
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 33.585913000000005,
            "dram": 10.885531,
            "package-0": 124.720994,
            "psys": 209.14826799999997,
            "uncore": 67.397411
          },
          "net_j": {
            "core": 15.91777577302597,
            "dram": 8.572365630825972,
            "package-0": 97.89517301921977,
            "psys": 161.57231947938268,
            "uncore": 67.38826215873642
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8333333333333334,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.552911,
            "dram": 0.303527,
            "package-0": 3.927052,
            "psys": 5.948593,
            "uncore": 0.737852
          },
          "logits": {
            "l2": 670.568131117269,
            "sha256": "f155fefe53ba4dfd4365752c28829b31e15f6eec90450cbf0ac8b52f5d1cfc31",
            "sum": -55835.01430409169,
            "top1": 1718
          },
          "net_j": {
            "core": 2.1162812743079327,
            "dram": 0.24636213040181287,
            "package-0": 3.2641099645249554,
            "psys": 4.772856526752401,
            "uncore": 0.7376259062132017
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.115762957,
          "watts": {
            "core": 22.05291801590728,
            "dram": 2.6219699968445,
            "package-0": 33.923217769912355,
            "psys": 51.38598005923432,
            "uncore": 6.373817835354706
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 874.4054818448777,
          "decode_tokens_per_s": 10.568191756125048,
          "end_to_end_ms": 4684.325619,
          "handoff_estimate_ms": 23.90276799999999,
          "handoff_plus_first_decode_ms": 115.762957,
          "output_tokens_per_s_e2e": 3.4156464134565536,
          "prefill_tokens_per_s": 1249.627824156777,
          "steady_tpot_ms": 91.860189,
          "steady_tpot_p95_ms": 100.3072394,
          "ttft_ms": 3264.972115
        },
        "prefill": {
          "gpu_active_fraction": 0.8978328173374613,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 323,
          "gross_j": {
            "core": 20.82795,
            "dram": 6.616072,
            "package-0": 91.281566,
            "psys": 150.646709,
            "uncore": 55.399883
          },
          "logits": {
            "l2": 662.3364776849841,
            "sha256": "d5055a398340419d05f270e2706e72a6086d083d704644e6d8b567ea276b1166",
            "sum": 8132.857491835253,
            "top1": 36
          },
          "net_j": {
            "core": 8.51326907003119,
            "dram": 5.003797140817782,
            "package-0": 72.58398504854449,
            "psys": 117.48630174000012,
            "uncore": 55.39350626318884
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.264972115,
          "watts": {
            "core": 6.379212215722094,
            "dram": 2.02637932789818,
            "package-0": 27.9578393887753,
            "psys": 46.140274309815965,
            "uncore": 16.967949816625005
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 706.6311889009895,
            "sha256": "b5b9231e8728eabecddc2a8f58f01d40f25e30ef87c52a6cee95dc6ef8f92e49",
            "sum": -130420.66216149065,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 10.205052,
            "dram": 3.965932,
            "package-0": 29.512376,
            "psys": 52.552966,
            "uncore": 11.259676
          },
          "net_j": {
            "core": 5.288225428686845,
            "dram": 3.322206359606377,
            "package-0": 22.04707800615032,
            "psys": 39.31316121263016,
            "uncore": 11.25712998933438
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.303590547,
          "token_latencies_s": [
            0.102212015,
            0.091879994,
            0.092006547,
            0.091841512,
            0.091815969,
            0.091823258,
            0.091361638,
            0.091772552,
            0.091610665,
            0.091878866,
            0.099281591,
            0.092324194,
            0.091923527,
            0.09183308
          ],
          "watts": {
            "core": 7.8284182280128185,
            "dram": 3.0423141753573946,
            "package-0": 22.639298871810553,
            "psys": 40.31401280175132,
            "uncore": 8.637432992984108
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.387475336901844,
    "compile_target": "NPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 399393,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 4096,
      "NPUW_LLM_MIN_RESPONSE_LEN": 4096,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05893102381378412,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 128.731178,
            "dram": 46.372623000000004,
            "package-0": 384.775261,
            "psys": 768.031237,
            "uncore": 134.87911
          },
          "net_j": {
            "core": 51.864434261035726,
            "dram": 36.308999009877816,
            "package-0": 268.0672300159029,
            "psys": 561.0479995430668,
            "uncore": 134.84290250755205
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 38,
          "gross_j": {
            "core": 5.061267,
            "dram": 0.532408,
            "package-0": 6.469527,
            "psys": 10.818393,
            "uncore": 0.0
          },
          "logits": {
            "l2": 913.718460033083,
            "sha256": "050909a6c91a93f5f72fdfebadeb65e20a07ce9afa0f13e33f5908e164cd68b6",
            "sum": -33577.34618644975,
            "top1": 4710
          },
          "net_j": {
            "core": 3.5794586640221815,
            "dram": 0.3384052207310998,
            "package-0": 4.21967323001223,
            "psys": 6.828247564505421,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 134433,
          "npu_utilization": 0.342182302461029,
          "seconds": 0.392869529,
          "watts": {
            "core": 12.882818916709624,
            "dram": 1.355177637103029,
            "package-0": 16.467367719933304,
            "psys": 27.536859444245675,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 200.98569672364553,
          "decode_tokens_per_s": 8.14839636520181,
          "end_to_end_ms": 20379.559674,
          "handoff_estimate_ms": 290.36244899999997,
          "handoff_plus_first_decode_ms": 392.869529,
          "output_tokens_per_s_e2e": 0.7851003778267404,
          "prefill_tokens_per_s": 220.08007756998083,
          "steady_tpot_ms": 102.50708,
          "steady_tpot_p95_ms": 108.01841569999999,
          "ttft_ms": 18538.706661
        },
        "prefill": {
          "gpu_active_fraction": 0.9989071038251366,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1830,
          "gross_j": {
            "core": 115.901254,
            "dram": 41.582779,
            "package-0": 356.59296,
            "psys": 717.111445,
            "uncore": 134.87911
          },
          "logits": {
            "l2": 2726.901179322358,
            "sha256": "13b2c92fc478d8b1d1227a16736da6fb25c66c7424aaab7c23a4956313737475",
            "sum": -678237.9385230131,
            "top1": 220
          },
          "net_j": {
            "core": 45.97776021711283,
            "dram": 32.42818606061,
            "package-0": 250.42697869288713,
            "psys": 528.8246721580113,
            "uncore": 134.84290250755205
          },
          "npu_busy_delta_us": 11648,
          "npu_utilization": 0.0006283070449841021,
          "seconds": 18.538706661,
          "watts": {
            "core": 6.25185219872011,
            "dram": 2.243024810758669,
            "package-0": 19.235050563163988,
            "psys": 38.68184863772574,
            "uncore": 7.275540439061269
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 844.4514378557863,
            "sha256": "b94aa7470bacc216bd3da9fb2a4994eff6fed67bd3d68a29a4cda171a3d91cd1",
            "sum": 14698.156317738816,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 143,
          "gross_j": {
            "core": 7.768657,
            "dram": 4.257436,
            "package-0": 21.712774,
            "psys": 40.101399,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.307215379900712,
            "dram": 3.5424077285367144,
            "package-0": 13.420578093003542,
            "psys": 25.39507982055013,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1430313,
          "npu_utilization": 0.9877964878776199,
          "seconds": 1.447983484,
          "token_latencies_s": [
            0.114787439,
            0.104373557,
            0.102584854,
            0.102702579,
            0.102476318,
            0.102493548,
            0.102288936,
            0.10253398,
            0.102520612,
            0.10205644,
            0.102034329,
            0.102316004,
            0.102579795,
            0.102189578
          ],
          "watts": {
            "core": 5.365155808641806,
            "dram": 2.9402517687832965,
            "package-0": 14.995180704699251,
            "psys": 27.69465221331212,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 123.883654,
            "dram": 46.325627,
            "package-0": 380.198673,
            "psys": 768.226855,
            "uncore": 135.395466
          },
          "net_j": {
            "core": 47.725245250188685,
            "dram": 36.3547403452083,
            "package-0": 264.5661185853531,
            "psys": 563.150989471775,
            "uncore": 135.3592353454075
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.852043,
            "dram": 0.389464,
            "package-0": 3.882436,
            "psys": 6.33818,
            "uncore": 0.0
          },
          "logits": {
            "l2": 928.4021349618147,
            "sha256": "dc9ab7845381a164647b92cacba6a85b45904ffc066588b6c31732bd1fd90879",
            "sum": 4324.664099577814,
            "top1": 1718
          },
          "net_j": {
            "core": 2.085854865644957,
            "dram": 0.28915235786738547,
            "package-0": 2.7191200564025593,
            "psys": 4.275023778780952,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 126123,
          "npu_utilization": 0.6208726910301483,
          "seconds": 0.203138263,
          "watts": {
            "core": 14.03991034421713,
            "dram": 1.9172360452840929,
            "package-0": 19.112283144805662,
            "psys": 31.201310409944778,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 202.85502676409223,
          "decode_tokens_per_s": 9.1396876314944,
          "end_to_end_ms": 20191.759925,
          "handoff_estimate_ms": 100.81320700000002,
          "handoff_plus_first_decode_ms": 203.13826300000002,
          "output_tokens_per_s_e2e": 0.7924024482972353,
          "prefill_tokens_per_s": 219.93938106248552,
          "steady_tpot_ms": 102.325056,
          "steady_tpot_p95_ms": 104.5934553,
          "ttft_ms": 18550.56598
        },
        "prefill": {
          "gpu_active_fraction": 0.9989077007099946,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1831,
          "gross_j": {
            "core": 112.40291,
            "dram": 41.694779,
            "package-0": 353.672862,
            "psys": 719.603505,
            "uncore": 135.395466
          },
          "logits": {
            "l2": 2748.16653849303,
            "sha256": "41fbb18485a75aa399f63fa651920f76f60f246ee8d8f1c1884ec0ebb413bcc3",
            "sum": -685130.2997232471,
            "top1": 220
          },
          "net_j": {
            "core": 42.434685749474,
            "dram": 32.53432981400291,
            "package-0": 247.43896569416086,
            "psys": 531.1962840089875,
            "uncore": 135.3592353454075
          },
          "npu_busy_delta_us": 11830,
          "npu_utilization": 0.0006377163916591186,
          "seconds": 18.55056598,
          "watts": {
            "core": 6.059271189956438,
            "dram": 2.2476284036267447,
            "package-0": 19.065340776195555,
            "psys": 38.791458210807654,
            "uncore": 7.298724262428138
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 931.1000582231685,
            "sha256": "0405a79d99d43e746aae0ab37db1bb0fa2c2f655e168c01217b32984b173a4fe",
            "sum": 51959.973022663034,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 142,
          "gross_j": {
            "core": 8.628701,
            "dram": 4.241384,
            "package-0": 22.643375,
            "psys": 42.28517,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.2047046350697306,
            "dram": 3.5312581733380046,
            "package-0": 14.40803283478964,
            "psys": 27.679681684006567,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1423531,
          "npu_utilization": 0.989899777747271,
          "seconds": 1.438055682,
          "token_latencies_s": [
            0.10803081,
            0.102742572,
            0.102346792,
            0.10263592,
            0.102321815,
            0.102043492,
            0.102112601,
            0.101988104,
            0.10227752,
            0.10246868,
            0.102328297,
            0.102100183,
            0.102135059,
            0.102470484
          ],
          "watts": {
            "core": 6.000255141719888,
            "dram": 2.949387880517411,
            "package-0": 15.745826314950717,
            "psys": 29.40440382752857,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 125.19163499999999,
            "dram": 46.294376,
            "package-0": 382.10504,
            "psys": 772.525121,
            "uncore": 136.046954
          },
          "net_j": {
            "core": 48.81744977781478,
            "dram": 36.29523924408186,
            "package-0": 266.14486864727934,
            "psys": 566.8682224964397,
            "uncore": 136.0106195597234
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 2.849114,
            "dram": 0.393981,
            "package-0": 3.859914,
            "psys": 6.339278,
            "uncore": 0.0
          },
          "logits": {
            "l2": 949.3787553822775,
            "sha256": "d8dc07559dcd42c867154c906a223ecc99445f3647bcf48618c4d6fa20fd761c",
            "sum": 19532.202365728095,
            "top1": 1718
          },
          "net_j": {
            "core": 2.0728924757012868,
            "dram": 0.29235575644721523,
            "package-0": 2.6813641967714297,
            "psys": 4.249104327121218,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 131611,
          "npu_utilization": 0.6395142041920637,
          "seconds": 0.2057984,
          "watts": {
            "core": 13.844198983082475,
            "dram": 1.9144026386988433,
            "package-0": 18.755801794377412,
            "psys": 30.80333957892773,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 202.28190978809923,
          "decode_tokens_per_s": 9.117084489125128,
          "end_to_end_ms": 20248.968404000003,
          "handoff_estimate_ms": 103.5047355,
          "handoff_plus_first_decode_ms": 205.7984,
          "output_tokens_per_s_e2e": 0.7901637101097626,
          "prefill_tokens_per_s": 219.31114628148657,
          "steady_tpot_ms": 102.29366449999999,
          "steady_tpot_p95_ms": 105.2186293,
          "ttft_ms": 18603.705599
        },
        "prefill": {
          "gpu_active_fraction": 0.9989136338946225,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1841,
          "gross_j": {
            "core": 114.189466,
            "dram": 41.656998,
            "package-0": 356.117558,
            "psys": 725.548374,
            "uncore": 136.046954
          },
          "logits": {
            "l2": 2756.8856185329796,
            "sha256": "69616f1e9441c7b5b9cc435f33c7c8f0372778126e2b065fc191ee8537c523a8",
            "sum": -687922.0755873509,
            "top1": 220
          },
          "net_j": {
            "core": 44.02081202766452,
            "dram": 32.47030795536078,
            "package-0": 249.57934597782827,
            "psys": 536.6014450587769,
            "uncore": 136.0106195597234
          },
          "npu_busy_delta_us": 9123,
          "npu_utilization": 0.0004903861734132357,
          "seconds": 18.603705599,
          "watts": {
            "core": 6.137995755326186,
            "dram": 2.2391774465749004,
            "package-0": 19.142291631358766,
            "psys": 39.00020725112959,
            "uncore": 7.312895448491342
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 994.7934300990154,
            "sha256": "0f720c6fde57b284973add3ed988f92a790012d4227c244f5e98a893d0af1f6e",
            "sum": 63108.52525581047,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 142,
          "gross_j": {
            "core": 8.153055,
            "dram": 4.243397,
            "package-0": 22.127568,
            "psys": 40.637469,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.7237452744489747,
            "dram": 3.532575532273869,
            "package-0": 13.884158472679665,
            "psys": 26.017673110541573,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1426629,
          "npu_utilization": 0.99108320778519,
          "seconds": 1.439464405,
          "token_latencies_s": [
            0.109740327,
            0.102783869,
            0.102348338,
            0.10224748,
            0.102184676,
            0.10277247,
            0.102080464,
            0.102401105,
            0.102170907,
            0.102339849,
            0.1023559,
            0.102009959,
            0.101968036,
            0.102022179
          ],
          "watts": {
            "core": 5.663950405220336,
            "dram": 2.9478999169833586,
            "package-0": 15.372084174599648,
            "psys": 28.23096483584115,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.5447567850351334,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 399235,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05975502822548151,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 32.33439,
            "dram": 10.583469,
            "package-0": 122.42882399999999,
            "psys": 203.42898300000002,
            "uncore": 67.017651
          },
          "net_j": {
            "core": 15.058093561304554,
            "dram": 8.321604598779293,
            "package-0": 96.19794128035583,
            "psys": 156.9081653588825,
            "uncore": 67.00870506010212
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.697381,
            "dram": 0.312988,
            "package-0": 4.187428,
            "psys": 6.295333,
            "uncore": 0.818418
          },
          "logits": {
            "l2": 670.568131117269,
            "sha256": "f155fefe53ba4dfd4365752c28829b31e15f6eec90450cbf0ac8b52f5d1cfc31",
            "sum": -55835.01430409169,
            "top1": 1718
          },
          "net_j": {
            "core": 2.2101573730297734,
            "dram": 0.24919922552578705,
            "package-0": 3.4476684188509785,
            "psys": 4.983359592901106,
            "uncore": 0.8181657078898723
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.129176839,
          "watts": {
            "core": 20.88130520053986,
            "dram": 2.422942087938845,
            "package-0": 32.416244525073104,
            "psys": 48.73422394242052,
            "uncore": 6.335640400675852
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 894.2377262438449,
          "decode_tokens_per_s": 10.384095135058907,
          "end_to_end_ms": 4580.437483,
          "handoff_estimate_ms": 36.29455049999999,
          "handoff_plus_first_decode_ms": 129.176839,
          "output_tokens_per_s_e2e": 3.493116118140019,
          "prefill_tokens_per_s": 1301.0533199167141,
          "steady_tpot_ms": 92.88228849999999,
          "steady_tpot_p95_ms": 100.38342,
          "ttft_ms": 3135.920671
        },
        "prefill": {
          "gpu_active_fraction": 0.9129032258064517,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 310,
          "gross_j": {
            "core": 19.354565,
            "dram": 6.297408,
            "package-0": 88.555437,
            "psys": 144.900325,
            "uncore": 54.865155
          },
          "logits": {
            "l2": 662.3364776849841,
            "sha256": "d5055a398340419d05f270e2706e72a6086d083d704644e6d8b567ea276b1166",
            "sum": 8132.857491835253,
            "top1": 36
          },
          "net_j": {
            "core": 7.526634737551589,
            "dram": 4.748859994026255,
            "package-0": 70.59689752764523,
            "psys": 113.05061758361306,
            "uncore": 54.859030310393074
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.135920671,
          "watts": {
            "core": 6.171892413920059,
            "dram": 2.008152839526979,
            "package-0": 28.239055221942504,
            "psys": 46.20662963192668,
            "uncore": 17.495708838356645
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 706.6311889009895,
            "sha256": "b5b9231e8728eabecddc2a8f58f01d40f25e30ef87c52a6cee95dc6ef8f92e49",
            "sum": -130420.66216149065,
            "top1": 42
          },
          "gpu_active_fraction": 0.9923076923076923,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 130,
          "gross_j": {
            "core": 10.282444,
            "dram": 3.973073,
            "package-0": 29.685959,
            "psys": 52.233325,
            "uncore": 11.334078
          },
          "net_j": {
            "core": 5.321301450723192,
            "dram": 3.3235453792272516,
            "package-0": 22.15337533385963,
            "psys": 38.874188182368336,
            "uncore": 11.331509041819166
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.315339973,
          "token_latencies_s": [
            0.102089293,
            0.093226358,
            0.099464873,
            0.092745762,
            0.093175623,
            0.092567746,
            0.09278905,
            0.09293055,
            0.092304634,
            0.093116181,
            0.092834027,
            0.092379771,
            0.093125818,
            0.092559209
          ],
          "watts": {
            "core": 7.8173279996562535,
            "dram": 3.020567367794881,
            "package-0": 22.56903888680041,
            "psys": 39.710893055935436,
            "uncore": 8.616842970376299
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 31.975198,
            "dram": 10.745334,
            "package-0": 122.58867599999999,
            "psys": 205.40725,
            "uncore": 67.15998499999999
          },
          "net_j": {
            "core": 14.39978398377925,
            "dram": 8.444308230902525,
            "package-0": 95.90363818274609,
            "psys": 158.08098228301205,
            "uncore": 67.15088417228338
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.796623,
            "dram": 0.302306,
            "package-0": 4.156789,
            "psys": 6.949201,
            "uncore": 0.732236
          },
          "logits": {
            "l2": 670.568131117269,
            "sha256": "f155fefe53ba4dfd4365752c28829b31e15f6eec90450cbf0ac8b52f5d1cfc31",
            "sum": -55835.01430409169,
            "top1": 1718
          },
          "net_j": {
            "core": 2.3612751511334626,
            "dram": 0.2453089575497613,
            "package-0": 3.4957932590210485,
            "psys": 5.776916305798361,
            "uncore": 0.7320105699893228
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.115423095,
          "watts": {
            "core": 24.229319097707435,
            "dram": 2.6191118857105677,
            "package-0": 36.013494526377066,
            "psys": 60.20633045752239,
            "uncore": 6.343929696218941
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 879.0186126480612,
          "decode_tokens_per_s": 10.527165091603338,
          "end_to_end_ms": 4659.742059,
          "handoff_estimate_ms": 22.689726500000006,
          "handoff_plus_first_decode_ms": 115.423095,
          "output_tokens_per_s_e2e": 3.433666455656489,
          "prefill_tokens_per_s": 1261.2613076033124,
          "steady_tpot_ms": 92.7333685,
          "steady_tpot_p95_ms": 97.10337725,
          "ttft_ms": 3234.857024
        },
        "prefill": {
          "gpu_active_fraction": 0.896875,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 320,
          "gross_j": {
            "core": 19.519847,
            "dram": 6.479781,
            "package-0": 89.433853,
            "psys": 146.950368,
            "uncore": 55.144574
          },
          "logits": {
            "l2": 662.3364776849841,
            "sha256": "d5055a398340419d05f270e2706e72a6086d083d704644e6d8b567ea276b1166",
            "sum": 8132.857491835253,
            "top1": 36
          },
          "net_j": {
            "core": 7.318752870708037,
            "dram": 4.882377264487787,
            "package-0": 70.90873274100984,
            "psys": 114.09582206042123,
            "uncore": 55.13825608023308
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.234857024,
          "watts": {
            "core": 6.034222488097204,
            "dram": 2.0031120237850732,
            "package-0": 27.646926073230986,
            "psys": 45.42716012168332,
            "uncore": 17.046989585898928
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 706.6311889009895,
            "sha256": "b5b9231e8728eabecddc2a8f58f01d40f25e30ef87c52a6cee95dc6ef8f92e49",
            "sum": -130420.66216149065,
            "top1": 42
          },
          "gpu_active_fraction": 0.9923076923076923,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 130,
          "gross_j": {
            "core": 9.658728,
            "dram": 3.963247,
            "package-0": 28.998034,
            "psys": 51.507681,
            "uncore": 11.283175
          },
          "net_j": {
            "core": 4.71975596193775,
            "dram": 3.3166220088649765,
            "package-0": 21.49911218271521,
            "psys": 38.20824391679247,
            "uncore": 11.280617522060982
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.30946194,
          "token_latencies_s": [
            0.104420762,
            0.093163247,
            0.092847174,
            0.092836554,
            0.092812207,
            0.092546024,
            0.092616328,
            0.092965256,
            0.09265453,
            0.092817805,
            0.092487615,
            0.092370358,
            0.092244352,
            0.092647497
          ],
          "watts": {
            "core": 7.37610441736092,
            "dram": 3.0266225225301318,
            "package-0": 22.144999494983413,
            "psys": 39.334996632280884,
            "uncore": 8.616649827943835
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.197904,
            "dram": 10.775973,
            "package-0": 128.108802,
            "psys": 212.037726,
            "uncore": 67.417186
          },
          "net_j": {
            "core": 19.638541837883924,
            "dram": 8.477048787651206,
            "package-0": 101.44813597469508,
            "psys": 164.7546819785683,
            "uncore": 67.40809348418765
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 12,
          "gross_j": {
            "core": 2.517145,
            "dram": 0.300292,
            "package-0": 3.875844,
            "psys": 5.881333,
            "uncore": 0.728391
          },
          "logits": {
            "l2": 670.568131117269,
            "sha256": "f155fefe53ba4dfd4365752c28829b31e15f6eec90450cbf0ac8b52f5d1cfc31",
            "sum": -55835.01430409169,
            "top1": 1718
          },
          "net_j": {
            "core": 2.0826394938183435,
            "dram": 0.2434052395734184,
            "package-0": 3.216127201661056,
            "psys": 4.711316527495859,
            "uncore": 0.7281660061677051
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.115199766,
          "watts": {
            "core": 21.850261397232355,
            "dram": 2.6067066837618404,
            "package-0": 33.64454750715379,
            "psys": 51.05334154932224,
            "uncore": 6.322851384958542
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 879.8221656698196,
          "decode_tokens_per_s": 10.586823056760185,
          "end_to_end_ms": 4655.486255999999,
          "handoff_estimate_ms": 23.738090000000003,
          "handoff_plus_first_decode_ms": 115.199766,
          "output_tokens_per_s_e2e": 3.4368053346477327,
          "prefill_tokens_per_s": 1259.7917110035642,
          "steady_tpot_ms": 91.461676,
          "steady_tpot_p95_ms": 101.17716695,
          "ttft_ms": 3238.630612
        },
        "prefill": {
          "gpu_active_fraction": 0.897196261682243,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 321,
          "gross_j": {
            "core": 22.002263,
            "dram": 6.509139,
            "package-0": 92.24189,
            "psys": 151.165323,
            "uncore": 55.40324
          },
          "logits": {
            "l2": 662.3364776849841,
            "sha256": "d5055a398340419d05f270e2706e72a6086d083d704644e6d8b567ea276b1166",
            "sum": 8132.857491835253,
            "top1": 36
          },
          "net_j": {
            "core": 9.786935814331333,
            "dram": 4.909871830177473,
            "package-0": 73.69515945932221,
            "psys": 118.27245093985911,
            "uncore": 55.39691471013108
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 3.238630612,
          "watts": {
            "core": 6.793693272235394,
            "dram": 2.0098429798946147,
            "package-0": 28.48175696796631,
            "psys": 46.67569139867069,
            "uncore": 17.10699571439733
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 706.6311889009895,
            "sha256": "b5b9231e8728eabecddc2a8f58f01d40f25e30ef87c52a6cee95dc6ef8f92e49",
            "sum": -130420.66216149065,
            "top1": 42
          },
          "gpu_active_fraction": 0.9922480620155039,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 129,
          "gross_j": {
            "core": 12.678496,
            "dram": 3.966542,
            "package-0": 31.991068,
            "psys": 54.99107,
            "uncore": 11.285555
          },
          "net_j": {
            "core": 7.7689665297342465,
            "dram": 3.3237717179003146,
            "package-0": 24.536849313711805,
            "psys": 41.77091451121332,
            "uncore": 11.283012767888867
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.301655878,
          "token_latencies_s": [
            0.105705234,
            0.092246253,
            0.091822925,
            0.091223243,
            0.091637262,
            0.098738977,
            0.091638716,
            0.091076312,
            0.090928838,
            0.091139372,
            0.09125447,
            0.09129244,
            0.091312382,
            0.09161097
          ],
          "watts": {
            "core": 9.74028252342744,
            "dram": 3.0473046425254955,
            "package-0": 24.577208570021146,
            "psys": 42.2470108493606,
            "uncore": 8.670152527056771
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.785680193454027,
    "compile_target": "AUTO:GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 399811,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.06051105260848999,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 59.252901,
            "dram": 19.136485999999998,
            "package-0": 247.554787,
            "psys": 406.515866,
            "uncore": 145.329156
          },
          "net_j": {
            "core": 25.151504512799242,
            "dram": 14.67182917844012,
            "package-0": 195.77808040052733,
            "psys": 314.689191636203,
            "uncore": 145.3114977568406
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8571428571428571,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 14,
          "gross_j": {
            "core": 2.945855,
            "dram": 0.32849,
            "package-0": 4.493031,
            "psys": 6.787702,
            "uncore": 0.846067
          },
          "logits": {
            "l2": 641.8176091039514,
            "sha256": "ae4c116f8db1f114b722fdb50e703656b8ec36d44932a98d79948cfafc69699b",
            "sum": -12608.67373151111,
            "top1": 78
          },
          "net_j": {
            "core": 2.4178898663994683,
            "dram": 0.25936722562446657,
            "package-0": 3.6914129246978282,
            "psys": 5.366021734889127,
            "uncore": 0.8457936112925225
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.139978571,
          "watts": {
            "core": 21.045042672995997,
            "dram": 2.3467163413177006,
            "package-0": 32.09799162758991,
            "psys": 48.49100795578204,
            "uncore": 6.04426087475918
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 906.0694069260272,
          "decode_tokens_per_s": 9.686000326255487,
          "end_to_end_ms": 9041.249972,
          "handoff_estimate_ms": 40.392893,
          "handoff_plus_first_decode_ms": 139.978571,
          "output_tokens_per_s_e2e": 1.769666810402397,
          "prefill_tokens_per_s": 1091.206623799312,
          "steady_tpot_ms": 99.585678,
          "steady_tpot_p95_ms": 107.27231370000001,
          "ttft_ms": 7492.62314
        },
        "prefill": {
          "gpu_active_fraction": 0.9284750337381916,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 741,
          "gross_j": {
            "core": 45.050178,
            "dram": 14.695336,
            "package-0": 211.082345,
            "psys": 343.780187,
            "uncore": 132.189908
          },
          "logits": {
            "l2": 630.9416903237038,
            "sha256": "14983026d2c3865bd74d1e89295c4c951cf5e5701b07ca9e53c79ce2d506866a",
            "sum": 80776.56900331145,
            "top1": 198
          },
          "net_j": {
            "core": 16.7898253698211,
            "dram": 10.995406110109062,
            "package-0": 168.17419057648246,
            "psys": 267.6820072861132,
            "uncore": 132.17527434185342
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 7.49262314,
          "watts": {
            "core": 6.012604285339781,
            "dram": 1.9613072385220751,
            "package-0": 28.172022141767563,
            "psys": 45.882487424824625,
            "uncore": 17.642674071553532
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 572.7403064888838,
            "sha256": "1f72cc218133f7bc0627789756871cfa6b42dada579c1e1db76b598cec8f6f75",
            "sum": -71473.40795485722,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 11.256868,
            "dram": 4.11266,
            "package-0": 31.979411,
            "psys": 55.947977,
            "uncore": 12.293181
          },
          "net_j": {
            "core": 5.943789276578674,
            "dram": 3.417055842706592,
            "package-0": 23.91247689934704,
            "psys": 41.6411626152007,
            "uncore": 12.290429803694638
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.408648261,
          "token_latencies_s": [
            0.109898315,
            0.099871449,
            0.099669116,
            0.09866253,
            0.099311326,
            0.099029114,
            0.099677494,
            0.09904256,
            0.105858313,
            0.099663692,
            0.099588846,
            0.09958251,
            0.099226278,
            0.099535382
          ],
          "watts": {
            "core": 7.991255384086263,
            "dram": 2.9195790843346665,
            "package-0": 22.702197479232893,
            "psys": 39.71749268357632,
            "uncore": 8.726934423837692
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 61.178005,
            "dram": 19.557933,
            "package-0": 250.28762899999998,
            "psys": 417.939299,
            "uncore": 145.322382
          },
          "net_j": {
            "core": 26.25683161547912,
            "dram": 14.985948535182136,
            "package-0": 197.26624176389998,
            "psys": 323.90516705512033,
            "uncore": 145.30429926338044
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.570673,
            "dram": 0.318481,
            "package-0": 4.025747,
            "psys": 6.196822,
            "uncore": 0.782835
          },
          "logits": {
            "l2": 641.8176091039514,
            "sha256": "ae4c116f8db1f114b722fdb50e703656b8ec36d44932a98d79948cfafc69699b",
            "sum": -12608.67373151111,
            "top1": 78
          },
          "net_j": {
            "core": 2.0807812737255413,
            "dram": 0.25434290998489095,
            "package-0": 3.2819364001513387,
            "psys": 4.877664057626371,
            "uncore": 0.7825813263057984
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.129884228,
          "watts": {
            "core": 19.79203356392125,
            "dram": 2.452037517596055,
            "package-0": 30.994887231419664,
            "psys": 47.71035017431062,
            "uncore": 6.027175216378081
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 884.7993665700598,
          "decode_tokens_per_s": 9.855231543392565,
          "end_to_end_ms": 9258.59614,
          "handoff_estimate_ms": 32.16617049999999,
          "handoff_plus_first_decode_ms": 129.88422799999998,
          "output_tokens_per_s_e2e": 1.728123762832148,
          "prefill_tokens_per_s": 1056.8001809501457,
          "steady_tpot_ms": 97.7180575,
          "steady_tpot_p95_ms": 107.81119354999998,
          "ttft_ms": 7736.561885
        },
        "prefill": {
          "gpu_active_fraction": 0.9137254901960784,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 765,
          "gross_j": {
            "core": 48.498899,
            "dram": 15.138999,
            "package-0": 215.589353,
            "psys": 356.844668,
            "uncore": 132.32565
          },
          "logits": {
            "l2": 630.9416903237038,
            "sha256": "14983026d2c3865bd74d1e89295c4c951cf5e5701b07ca9e53c79ce2d506866a",
            "sum": 80776.56900331145,
            "top1": 198
          },
          "net_j": {
            "core": 19.31846873812223,
            "dram": 11.318609794717954,
            "package-0": 171.28422969859236,
            "psys": 278.2689454937781,
            "uncore": 132.3105399110846
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 7.736561885,
          "watts": {
            "core": 6.26879222591522,
            "dram": 1.956812241023003,
            "package-0": 27.86629981180587,
            "psys": 46.124450796660305,
            "uncore": 17.103934792605873
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 572.7403064888838,
            "sha256": "1f72cc218133f7bc0627789756871cfa6b42dada579c1e1db76b598cec8f6f75",
            "sum": -71473.40795485722,
            "top1": 42
          },
          "gpu_active_fraction": 0.9927536231884058,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 138,
          "gross_j": {
            "core": 10.108433,
            "dram": 4.100453,
            "package-0": 30.672529,
            "psys": 54.897809,
            "uncore": 12.213897
          },
          "net_j": {
            "core": 4.857581603631346,
            "dram": 3.41299583047929,
            "package-0": 22.700075665156294,
            "psys": 40.75855750371584,
            "uncore": 12.211178025990067
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.392150027,
          "token_latencies_s": [
            0.113735042,
            0.098577988,
            0.097677854,
            0.097593822,
            0.097758261,
            0.097606737,
            0.097638762,
            0.097464358,
            0.097638063,
            0.097637551,
            0.104621429,
            0.098021256,
            0.098232888,
            0.097920396
          ],
          "watts": {
            "core": 7.261022737458166,
            "dram": 2.9454102794051806,
            "package-0": 22.03248816946652,
            "psys": 39.4338310780351,
            "uncore": 8.773405712831265
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 59.900603000000004,
            "dram": 19.610851,
            "package-0": 249.049716,
            "psys": 412.77097599999996,
            "uncore": 145.30926
          },
          "net_j": {
            "core": 25.00429956929731,
            "dram": 15.042122583919257,
            "package-0": 196.06608922049801,
            "psys": 318.80381272461636,
            "uncore": 145.2911901414364
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7857142857142857,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 14,
          "gross_j": {
            "core": 2.978203,
            "dram": 0.320922,
            "package-0": 4.455494,
            "psys": 6.617048,
            "uncore": 0.798399
          },
          "logits": {
            "l2": 641.8176091039514,
            "sha256": "ae4c116f8db1f114b722fdb50e703656b8ec36d44932a98d79948cfafc69699b",
            "sum": -12608.67373151111,
            "top1": 78
          },
          "net_j": {
            "core": 2.4825701657042822,
            "dram": 0.25603226695173686,
            "package-0": 3.702966582185124,
            "psys": 5.282430665605476,
            "uncore": 0.7981423534691439
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.131406359,
          "watts": {
            "core": 22.664070617769724,
            "dram": 2.4422105782567187,
            "package-0": 33.90622823664112,
            "psys": 50.35561482987288,
            "uncore": 6.075801856742716
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 885.4299468098451,
          "decode_tokens_per_s": 9.70446882113418,
          "end_to_end_ms": 9252.002408,
          "handoff_estimate_ms": 31.965154999999996,
          "handoff_plus_first_decode_ms": 131.406359,
          "output_tokens_per_s_e2e": 1.7293553648629787,
          "prefill_tokens_per_s": 1060.9469989685113,
          "steady_tpot_ms": 99.441204,
          "steady_tpot_p95_ms": 107.64871504999999,
          "ttft_ms": 7706.322755
        },
        "prefill": {
          "gpu_active_fraction": 0.9145860709592641,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 761,
          "gross_j": {
            "core": 45.479925,
            "dram": 15.16677,
            "package-0": 212.387945,
            "psys": 349.058188,
            "uncore": 132.189909
          },
          "logits": {
            "l2": 630.9416903237038,
            "sha256": "14983026d2c3865bd74d1e89295c4c951cf5e5701b07ca9e53c79ce2d506866a",
            "sum": 80776.56900331145,
            "top1": 198
          },
          "net_j": {
            "core": 16.413549383746624,
            "dram": 11.361313170047763,
            "package-0": 168.25599272767374,
            "psys": 270.7895866055883,
            "uncore": 132.17485797038637
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 7.706322755,
          "watts": {
            "core": 5.901637713070324,
            "dram": 1.9680943145236847,
            "package-0": 27.560219283859983,
            "psys": 45.29503877494941,
            "uncore": 17.153435328702372
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 572.7403064888838,
            "sha256": "1f72cc218133f7bc0627789756871cfa6b42dada579c1e1db76b598cec8f6f75",
            "sum": -71473.40795485722,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928571428571429,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 11.442475,
            "dram": 4.123159,
            "package-0": 32.206277,
            "psys": 57.09574,
            "uncore": 12.320952
          },
          "net_j": {
            "core": 6.108180019846403,
            "dram": 3.424777146919757,
            "package-0": 24.107129910639138,
            "psys": 42.7317954534226,
            "uncore": 12.318189817580903
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.414273294,
          "token_latencies_s": [
            0.111278265,
            0.099954819,
            0.105694342,
            0.099467568,
            0.099617773,
            0.105665746,
            0.099892691,
            0.09941484,
            0.099150481,
            0.098806237,
            0.09911801,
            0.098637142,
            0.098750954,
            0.098797943
          ],
          "watts": {
            "core": 8.090709941667045,
            "dram": 2.9153905525136787,
            "package-0": 22.772315037435757,
            "psys": 40.37107979216356,
            "uncore": 8.711860750161348
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 11.432788781821728,
    "compile_target": "NPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 399882,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 8192,
      "NPUW_LLM_MIN_RESPONSE_LEN": 8192,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.058648111298680305,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 243.79735000000002,
            "dram": 91.704476,
            "package-0": 769.0989219999999,
            "psys": 1562.7052279999998,
            "uncore": 289.146354
          },
          "net_j": {
            "core": 90.593715480322,
            "dram": 71.64659795378061,
            "package-0": 536.487344422071,
            "psys": 1150.1655167987642,
            "uncore": 289.07095152272984
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 46,
          "gross_j": {
            "core": 6.469832,
            "dram": 0.63281,
            "package-0": 8.125956,
            "psys": 13.474514,
            "uncore": 0.0
          },
          "logits": {
            "l2": 918.8020555783207,
            "sha256": "1bb6018b81cc8c62cfc72ed7206d2c94b7a2abbf7d1eaf57f9ec872a0d5d3e65",
            "sum": -25677.505140628666,
            "top1": 1825
          },
          "net_j": {
            "core": 4.696840984083592,
            "dram": 0.4006847133787733,
            "package-0": 5.433994816854322,
            "psys": 8.700285216448568,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 145249,
          "npu_utilization": 0.3089941829088565,
          "seconds": 0.470070338,
          "watts": {
            "core": 13.763540212996807,
            "dram": 1.3462027889111352,
            "package-0": 17.286681041338117,
            "psys": 28.664888870312,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 201.68080337897354,
          "decode_tokens_per_s": 7.45678042608262,
          "end_to_end_ms": 40618.640261,
          "handoff_estimate_ms": 361.84099299999997,
          "handoff_plus_first_decode_ms": 470.070338,
          "output_tokens_per_s_e2e": 0.3939078190995577,
          "prefill_tokens_per_s": 211.77480207661233,
          "steady_tpot_ms": 108.22934500000001,
          "steady_tpot_p95_ms": 117.97439245,
          "ttft_ms": 38607.048241000004
        },
        "prefill": {
          "gpu_active_fraction": 0.9992128050380478,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3811,
          "gross_j": {
            "core": 228.675562,
            "dram": 86.60067,
            "package-0": 737.592777,
            "psys": 1506.485767,
            "uncore": 289.146354
          },
          "logits": {
            "l2": 2756.7075776903853,
            "sha256": "84cd27dc25b56e5e390e1b1b887e9cf001acecb3b20c00fa1bc22fbf10010345",
            "sum": -687907.677328907,
            "top1": 220
          },
          "net_j": {
            "core": 83.05916345509891,
            "dram": 67.53613558017099,
            "package-0": 516.5010235769098,
            "psys": 1114.3766163899886,
            "uncore": 289.07095152272984
          },
          "npu_busy_delta_us": 11610,
          "npu_utilization": 0.00030072229110927954,
          "seconds": 38.607048241,
          "watts": {
            "core": 5.923155807523006,
            "dram": 2.2431310847544053,
            "package-0": 19.105132627484572,
            "psys": 39.02100356380363,
            "uncore": 7.489470632280343
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 997.8621615342367,
            "sha256": "29494008627b5f3ecc3f195d249808708e3e02693b135cdb6ada818d59389de2",
            "sum": 88720.19780683517,
            "top1": 220
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 151,
          "gross_j": {
            "core": 8.651956,
            "dram": 4.470996,
            "package-0": 23.380189,
            "psys": 42.744947,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.837711041139503,
            "dram": 3.709777660230846,
            "package-0": 14.552326028306952,
            "psys": 27.088615192326966,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1515885,
          "npu_utilization": 0.9833692368395762,
          "seconds": 1.541521682,
          "token_latencies_s": [
            0.13144969,
            0.110718463,
            0.108562676,
            0.108409888,
            0.108237745,
            0.10857501,
            0.108192256,
            0.108145068,
            0.108166442,
            0.108141932,
            0.108181916,
            0.108220945,
            0.10806269,
            0.108403835
          ],
          "watts": {
            "core": 5.612607400224683,
            "dram": 2.9003782770017508,
            "package-0": 15.166954362695758,
            "psys": 27.729059862811585,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 220.631943,
            "dram": 91.64277100000001,
            "package-0": 750.822663,
            "psys": 1555.000536,
            "uncore": 289.427665
          },
          "net_j": {
            "core": 67.98900070301652,
            "dram": 71.6583004565434,
            "package-0": 519.0623935706379,
            "psys": 1143.9706310664053,
            "uncore": 289.35160010883976
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 3.697134,
            "dram": 0.481139,
            "package-0": 4.926074,
            "psys": 8.250284,
            "uncore": 0.003113
          },
          "logits": {
            "l2": 928.0207259122993,
            "sha256": "96bb7937c1ddec1e5746531cc2c98ec321f0f8ff76d31787766875d670a08fba",
            "sum": -8464.025744725019,
            "top1": 1825
          },
          "net_j": {
            "core": 2.7011726789501522,
            "dram": 0.3507447650394482,
            "package-0": 3.41388967864014,
            "psys": 5.5684050939174465,
            "uncore": 0.0025972754550806203
          },
          "npu_busy_delta_us": 139865,
          "npu_utilization": 0.5296759625640716,
          "seconds": 0.264057669,
          "watts": {
            "core": 14.0012369797902,
            "dram": 1.822098187195616,
            "package-0": 18.655296089885578,
            "psys": 31.244250664047183,
            "uncore": 0.01178909142002613
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 202.42162281169493,
          "decode_tokens_per_s": 8.390060790330883,
          "end_to_end_ms": 40469.984808,
          "handoff_estimate_ms": 155.709177,
          "handoff_plus_first_decode_ms": 264.057669,
          "output_tokens_per_s_e2e": 0.39535473205409166,
          "prefill_tokens_per_s": 211.36361147807168,
          "steady_tpot_ms": 108.348492,
          "steady_tpot_p95_ms": 111.32695169999998,
          "ttft_ms": 38682.155092
        },
        "prefill": {
          "gpu_active_fraction": 0.998949027850762,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3806,
          "gross_j": {
            "core": 209.477858,
            "dram": 86.706382,
            "package-0": 723.630899,
            "psys": 1504.622125,
            "uncore": 289.424552
          },
          "logits": {
            "l2": 2763.2184793149227,
            "sha256": "0d0b42e7fc1606224918b1e28631651539a24fa3768396a1ab8522ae44b7e5fe",
            "sum": -690004.62714516,
            "top1": 220
          },
          "net_j": {
            "core": 63.57817467558411,
            "dram": 67.6047590895639,
            "package-0": 502.10902967480774,
            "psys": 1111.7501581454844,
            "uncore": 289.3490028333847
          },
          "npu_busy_delta_us": 8960,
          "npu_utilization": 0.00023163135504446208,
          "seconds": 38.682155092,
          "watts": {
            "core": 5.415361618342792,
            "dram": 2.2415085662570045,
            "package-0": 18.70709884904155,
            "psys": 38.897060451297776,
            "uncore": 7.482120665501829
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1053.16157417548,
            "sha256": "976fdf28a39fa930aa5a083ec7fdcd6fd44fbeebbe2fcfebf1c014f2501fccd0",
            "sum": 84184.23420181125,
            "top1": 220
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 150,
          "gross_j": {
            "core": 7.456951,
            "dram": 4.45525,
            "package-0": 22.26569,
            "psys": 42.128127,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.7096533484822505,
            "dram": 3.7027966019400527,
            "package-0": 13.539474217189998,
            "psys": 26.652067827003457,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1508805,
          "npu_utilization": 0.9901776338334417,
          "seconds": 1.523772047,
          "token_latencies_s": [
            0.115988844,
            0.108816702,
            0.108491134,
            0.10849455,
            0.108279471,
            0.108737424,
            0.108417513,
            0.108208627,
            0.108577581,
            0.107911744,
            0.107922596,
            0.107959962,
            0.107831916,
            0.108088828
          ],
          "watts": {
            "core": 4.893744451265682,
            "dram": 2.9238297216250224,
            "package-0": 14.612218437683415,
            "psys": 27.647263304863603,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 235.62830300000002,
            "dram": 91.64258799999999,
            "package-0": 761.199041,
            "psys": 1555.684311,
            "uncore": 289.522086
          },
          "net_j": {
            "core": 83.14381982333549,
            "dram": 71.67886339850902,
            "package-0": 529.6793626380402,
            "psys": 1145.08109750841,
            "uncore": 289.4466305862844
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 27,
          "gross_j": {
            "core": 3.603873,
            "dram": 0.482482,
            "package-0": 4.839954,
            "psys": 8.271402,
            "uncore": 0.0
          },
          "logits": {
            "l2": 937.7241911868175,
            "sha256": "f634f53d4b1a8bdc58612b054b7078b9f79e875e90249ee418cda8638c28c679",
            "sum": 8204.341230405495,
            "top1": 1825
          },
          "net_j": {
            "core": 2.589732372939451,
            "dram": 0.3497076759242601,
            "package-0": 3.3001677417560775,
            "psys": 5.54057069357129,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 136832,
          "npu_utilization": 0.5089008556758483,
          "seconds": 0.26887752,
          "watts": {
            "core": 13.403400180126626,
            "dram": 1.794430415752124,
            "package-0": 18.000590008417216,
            "psys": 30.762713074711492,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 202.63197570541928,
          "decode_tokens_per_s": 8.362041149715992,
          "end_to_end_ms": 40427.972789,
          "handoff_estimate_ms": 160.46092999999996,
          "handoff_plus_first_decode_ms": 268.87752,
          "output_tokens_per_s_e2e": 0.39576557754964703,
          "prefill_tokens_per_s": 211.62622949967678,
          "steady_tpot_ms": 108.41659000000001,
          "steady_tpot_p95_ms": 111.1662418,
          "ttft_ms": 38634.152389
        },
        "prefill": {
          "gpu_active_fraction": 0.9992128050380478,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3811,
          "gross_j": {
            "core": 225.904512,
            "dram": 86.696677,
            "package-0": 735.47529,
            "psys": 1506.959949,
            "uncore": 289.522086
          },
          "logits": {
            "l2": null,
            "sha256": "b0ba3595a28b4c7ae5c9e581e843cac4c1d56ce6f3684ebf9bc0b9924c710c7a",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 80.18588319929822,
            "dram": 67.61875828932233,
            "package-0": 514.2283187117948,
            "psys": 1114.575517452064,
            "uncore": 289.4466305862844
          },
          "npu_busy_delta_us": 8911,
          "npu_utilization": 0.00023065084773380868,
          "seconds": 38.634152389,
          "watts": {
            "core": 5.8472749634937,
            "dram": 2.244042424616114,
            "package-0": 19.03692056175163,
            "psys": 39.0059016651046,
            "uncore": 7.493941709523136
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1073.8044177318684,
            "sha256": "f44e69db153b06ae69976129b2fb2b8c01efb46054fd94323dfac83613b8fa55",
            "sum": 74667.62392079458,
            "top1": 220
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 151,
          "gross_j": {
            "core": 6.119918,
            "dram": 4.463429,
            "package-0": 20.883797,
            "psys": 40.45296,
            "uncore": 0.0
          },
          "net_j": {
            "core": 0.36820425109782384,
            "dram": 3.71039743326243,
            "package-0": 12.150876184489242,
            "psys": 24.965009362774687,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1508335,
          "npu_utilization": 0.9891091789615096,
          "seconds": 1.52494288,
          "token_latencies_s": [
            0.115011387,
            0.109095779,
            0.108347609,
            0.108177885,
            0.108474542,
            0.108210081,
            0.108682487,
            0.108365508,
            0.108171777,
            0.108375109,
            0.108572627,
            0.108611577,
            0.108458071,
            0.10833528
          ],
          "watts": {
            "core": 4.013211301396417,
            "dram": 2.926948319533122,
            "package-0": 13.694806063817945,
            "psys": 26.527524755550186,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 3.51444427203387,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 399723,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_3b_ov_fp16/openvino_model.xml",
      "key": "qwen25_3b_fp16",
      "size_b": 3.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.061123667284846306,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 54.979962,
            "dram": 19.41206,
            "package-0": 243.268175,
            "psys": 403.04382899999996,
            "uncore": 145.02385900000002
          },
          "net_j": {
            "core": 20.75228509830278,
            "dram": 14.930870168883676,
            "package-0": 191.29973478712247,
            "psys": 310.8771125354391,
            "uncore": 145.00613536684207
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.7857142857142857,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 14,
          "gross_j": {
            "core": 2.786309,
            "dram": 0.327819,
            "package-0": 4.329212,
            "psys": 6.560408,
            "uncore": 0.841734
          },
          "logits": {
            "l2": 641.8176091039514,
            "sha256": "ae4c116f8db1f114b722fdb50e703656b8ec36d44932a98d79948cfafc69699b",
            "sum": -12608.67373151111,
            "top1": 78
          },
          "net_j": {
            "core": 2.2585994859045027,
            "dram": 0.25872969209459495,
            "package-0": 3.527982035962369,
            "psys": 5.1394160553530295,
            "uncore": 0.8414607436563504
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.139910799,
          "watts": {
            "core": 19.914895918791803,
            "dram": 2.3430571645867024,
            "package-0": 30.942657971669508,
            "psys": 46.88993306370868,
            "uncore": 6.016218948188552
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 902.7265326609174,
          "decode_tokens_per_s": 9.73978923886958,
          "end_to_end_ms": 9074.730501,
          "handoff_estimate_ms": 41.047672000000006,
          "handoff_plus_first_decode_ms": 139.910799,
          "output_tokens_per_s_e2e": 1.7631377591033544,
          "prefill_tokens_per_s": 1085.1191989132783,
          "steady_tpot_ms": 98.86312699999999,
          "steady_tpot_p95_ms": 106.38641039999999,
          "ttft_ms": 7534.656108
        },
        "prefill": {
          "gpu_active_fraction": 0.9288590604026845,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 745,
          "gross_j": {
            "core": 41.544877,
            "dram": 14.973412,
            "package-0": 207.61336,
            "psys": 341.30418,
            "uncore": 131.899198
          },
          "logits": {
            "l2": 630.9416903237038,
            "sha256": "14983026d2c3865bd74d1e89295c4c951cf5e5701b07ca9e53c79ce2d506866a",
            "sum": 80776.56900331145,
            "top1": 198
          },
          "net_j": {
            "core": 13.12598623496796,
            "dram": 11.252725823324294,
            "package-0": 164.4644945400239,
            "psys": 264.77909607739946,
            "uncore": 131.88448224829645
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 7.534656108,
          "watts": {
            "core": 5.513838508951894,
            "dram": 1.987272117715077,
            "package-0": 27.554457300256125,
            "psys": 45.29791076166259,
            "uncore": 17.50566928462132
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 572.7403064888838,
            "sha256": "1f72cc218133f7bc0627789756871cfa6b42dada579c1e1db76b598cec8f6f75",
            "sum": -71473.40795485722,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928057553956835,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 139,
          "gross_j": {
            "core": 10.648776,
            "dram": 4.110829,
            "package-0": 31.325603,
            "psys": 55.179241,
            "uncore": 12.282927
          },
          "net_j": {
            "core": 5.367699377430318,
            "dram": 3.4194146534647865,
            "package-0": 23.3072582111362,
            "psys": 40.95860040268663,
            "uncore": 12.280192374889271
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.400163594,
          "token_latencies_s": [
            0.108406522,
            0.09943424,
            0.098498826,
            0.098789552,
            0.098829543,
            0.105298658,
            0.098887598,
            0.099213776,
            0.098774743,
            0.098893934,
            0.098722326,
            0.098463872,
            0.099079368,
            0.098838656
          ],
          "watts": {
            "core": 7.60537986106215,
            "dram": 2.93596335286518,
            "package-0": 22.37281638676859,
            "psys": 39.409138500997194,
            "uncore": 8.772494194703366
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 58.368625,
            "dram": 19.586986,
            "package-0": 248.11380599999998,
            "psys": 413.946328,
            "uncore": 145.851861
          },
          "net_j": {
            "core": 23.45702970533852,
            "dram": 15.016255527353522,
            "package-0": 195.10696133398565,
            "psys": 319.9379874956227,
            "uncore": 145.8337832230671
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.289179,
            "dram": 0.322448,
            "package-0": 3.777395,
            "psys": 5.901169,
            "uncore": 0.802427
          },
          "logits": {
            "l2": 641.8176091039514,
            "sha256": "ae4c116f8db1f114b722fdb50e703656b8ec36d44932a98d79948cfafc69699b",
            "sum": -12608.67373151111,
            "top1": 78
          },
          "net_j": {
            "core": 1.7911000305187375,
            "dram": 0.2572380116171292,
            "package-0": 3.021153575227899,
            "psys": 4.559964825115906,
            "uncore": 0.8021690868216058
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.132054899,
          "watts": {
            "core": 17.33505547567758,
            "dram": 2.441772341971198,
            "package-0": 28.604732036484307,
            "psys": 44.68724026664092,
            "uncore": 6.076465213153508
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 885.0421136507651,
          "decode_tokens_per_s": 9.861322321369133,
          "end_to_end_ms": 9256.056716,
          "handoff_estimate_ms": 34.0448985,
          "handoff_plus_first_decode_ms": 132.054899,
          "output_tokens_per_s_e2e": 1.7285978782241507,
          "prefill_tokens_per_s": 1057.0186944027648,
          "steady_tpot_ms": 98.0100005,
          "steady_tpot_p95_ms": 106.1401223,
          "ttft_ms": 7734.962535000001
        },
        "prefill": {
          "gpu_active_fraction": 0.912303664921466,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 764,
          "gross_j": {
            "core": 46.320377,
            "dram": 15.164085,
            "package-0": 213.901064,
            "psys": 353.616038,
            "uncore": 132.733486
          },
          "logits": {
            "l2": 630.9416903237038,
            "sha256": "14983026d2c3865bd74d1e89295c4c951cf5e5701b07ca9e53c79ce2d506866a",
            "sum": 80776.56900331145,
            "top1": 198
          },
          "net_j": {
            "core": 17.145979097435458,
            "dram": 11.344485569241353,
            "package-0": 169.6050997281465,
            "psys": 275.0565591539024,
            "uncore": 132.7183790347358
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 7.734962535,
          "watts": {
            "core": 5.9884423215244444,
            "dram": 1.9604600450724743,
            "package-0": 27.653794447240976,
            "psys": 45.716580578111355,
            "uncore": 17.160197660866885
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 572.7403064888838,
            "sha256": "1f72cc218133f7bc0627789756871cfa6b42dada579c1e1db76b598cec8f6f75",
            "sum": -71473.40795485722,
            "top1": 42
          },
          "gpu_active_fraction": 0.9927536231884058,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 138,
          "gross_j": {
            "core": 9.759069,
            "dram": 4.100453,
            "package-0": 30.435347,
            "psys": 54.429121,
            "uncore": 12.315948
          },
          "net_j": {
            "core": 4.519950577384324,
            "dram": 3.4145319464950394,
            "package-0": 22.480708030611247,
            "psys": 40.321463516604375,
            "uncore": 12.313235101509692
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.389039282,
          "token_latencies_s": [
            0.108032674,
            0.099217043,
            0.097602912,
            0.105121056,
            0.098332319,
            0.098163488,
            0.097957261,
            0.09808941,
            0.09806274,
            0.097918848,
            0.097561998,
            0.09749588,
            0.097647183,
            0.097783248
          ],
          "watts": {
            "core": 7.025768908384263,
            "dram": 2.95200650776124,
            "package-0": 21.91107724194657,
            "psys": 39.18472407895517,
            "uncore": 8.866522465993155
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 61.078823,
            "dram": 19.575633999999997,
            "package-0": 251.020964,
            "psys": 415.161291,
            "uncore": 146.055046
          },
          "net_j": {
            "core": 26.13484856081367,
            "dram": 15.000664352909933,
            "package-0": 197.96495755063657,
            "psys": 321.0657614222492,
            "uncore": 146.03695145663326
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.8461538461538461,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 13,
          "gross_j": {
            "core": 2.646721,
            "dram": 0.3208,
            "package-0": 4.119923,
            "psys": 6.784163,
            "uncore": 0.793516
          },
          "logits": {
            "l2": 641.8176091039514,
            "sha256": "ae4c116f8db1f114b722fdb50e703656b8ec36d44932a98d79948cfafc69699b",
            "sum": -12608.67373151111,
            "top1": 78
          },
          "net_j": {
            "core": 2.149070430433175,
            "dram": 0.2556460990152235,
            "package-0": 3.3643320218066894,
            "psys": 5.44411240073136,
            "uncore": 0.793258308653866
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.131941318,
          "watts": {
            "core": 20.059834478840052,
            "dram": 2.4313839278155456,
            "package-0": 31.225419470192044,
            "psys": 51.418032674192325,
            "uncore": 6.014158506435414
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 884.2220321640885,
          "decode_tokens_per_s": 9.705495964586772,
          "end_to_end_ms": 9264.641348,
          "handoff_estimate_ms": 33.072184500000006,
          "handoff_plus_first_decode_ms": 131.941318,
          "output_tokens_per_s_e2e": 1.7269961565704854,
          "prefill_tokens_per_s": 1059.1873700275985,
          "steady_tpot_ms": 98.8691335,
          "steady_tpot_p95_ms": 107.23170945000001,
          "ttft_ms": 7719.125276
        },
        "prefill": {
          "gpu_active_fraction": 0.9136125654450262,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 764,
          "gross_j": {
            "core": 47.70575,
            "dram": 15.140892,
            "package-0": 215.360533,
            "psys": 352.979565,
            "uncore": 132.884852
          },
          "logits": {
            "l2": 630.9416903237038,
            "sha256": "14983026d2c3865bd74d1e89295c4c951cf5e5701b07ca9e53c79ce2d506866a",
            "sum": 80776.56900331145,
            "top1": 198
          },
          "net_j": {
            "core": 18.5910863875062,
            "dram": 11.32911316115598,
            "package-0": 171.15526427523594,
            "psys": 274.5809359069246,
            "uncore": 132.8697759660972
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 7.719125276,
          "watts": {
            "core": 6.1802015505985946,
            "dram": 1.961477688032278,
            "package-0": 27.89960329696818,
            "psys": 45.72792283828715,
            "uncore": 17.215014298726352
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 572.7403064888838,
            "sha256": "1f72cc218133f7bc0627789756871cfa6b42dada579c1e1db76b598cec8f6f75",
            "sum": -71473.40795485722,
            "top1": 42
          },
          "gpu_active_fraction": 0.9928571428571429,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 140,
          "gross_j": {
            "core": 10.726352,
            "dram": 4.113942,
            "package-0": 31.540508,
            "psys": 55.397563,
            "uncore": 12.376678
          },
          "net_j": {
            "core": 5.394691742874296,
            "dram": 3.4159050927387296,
            "package-0": 23.44536125359393,
            "psys": 41.0407131145932,
            "uncore": 12.373917181882211
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.413574754,
          "token_latencies_s": [
            0.109016081,
            0.106270894,
            0.099583976,
            0.098923403,
            0.105318778,
            0.09895458,
            0.098755646,
            0.098749229,
            0.098366066,
            0.098175806,
            0.105128852,
            0.098814864,
            0.098753034,
            0.098734115
          ],
          "watts": {
            "core": 7.5881038265911185,
            "dram": 2.9103108897203747,
            "package-0": 22.312585811786505,
            "psys": 39.18969466824532,
            "uncore": 8.755587891604351
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 10.92683682218194,
    "compile_target": "AUTO:GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394304,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05495245289057493,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 24.365965,
            "dram": 12.233795,
            "package-0": 97.665644,
            "psys": 168.58648300000002,
            "uncore": 48.311949
          },
          "net_j": {
            "core": 14.7783067474509,
            "dram": 10.318297851833647,
            "package-0": 80.89885503160497,
            "psys": 129.8226451709811,
            "uncore": 48.30365924597621
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.95,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.811028,
            "dram": 0.633727,
            "package-0": 5.787522,
            "psys": 9.31919,
            "uncore": 1.68365
          },
          "logits": {
            "l2": 826.5092086937404,
            "sha256": "4dd5fd8e1442cc127c6d0044f29c595ddad6c04dc29eeda373180bd92ad1842c",
            "sum": 6180.617078131763,
            "top1": 78
          },
          "net_j": {
            "core": 2.3141447120341847,
            "dram": 0.5344557811059611,
            "package-0": 4.9185780919263,
            "psys": 7.310242383460212,
            "uncore": 1.683220380991138
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.202271215,
          "watts": {
            "core": 13.897320980644723,
            "dram": 3.133055783542903,
            "package-0": 28.612682234592796,
            "psys": 46.07274445847374,
            "uncore": 8.323725152884458
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 262.3660990401052,
          "decode_tokens_per_s": 5.34787218891652,
          "end_to_end_ms": 3902.943268,
          "handoff_estimate_ms": 16.331582499999996,
          "handoff_plus_first_decode_ms": 202.271215,
          "output_tokens_per_s_e2e": 4.099470297501644,
          "prefill_tokens_per_s": 917.9580468748337,
          "steady_tpot_ms": 185.93963250000002,
          "steady_tpot_p95_ms": 186.1239023,
          "ttft_ms": 1098.089399
        },
        "prefill": {
          "gpu_active_fraction": 0.9357798165137615,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 109,
          "gross_j": {
            "core": 5.378038,
            "dram": 2.915642,
            "package-0": 34.537327,
            "psys": 56.04368,
            "uncore": 22.934512
          },
          "logits": {
            "l2": 848.6664321595354,
            "sha256": "0a12f44f7c57878debefd7f9c698d7b450eec350fae83b41138241ea19af3147",
            "sum": -208252.0210649966,
            "top1": 290
          },
          "net_j": {
            "core": 2.6805594138575,
            "dram": 2.376718688082619,
            "package-0": 29.820006773385643,
            "psys": 45.137510870256754,
            "uncore": 22.932179685541712
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.098089399,
          "watts": {
            "core": 4.897632200891505,
            "dram": 2.6551954719307878,
            "package-0": 31.45219964007684,
            "psys": 51.03744745285534,
            "uncore": 20.88583317613833
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 760.3157061515358,
            "sha256": "779808bb2a5fa209d3cc4c2c6cffdb0cd54061013fde8ed3d471ca149f7ac757",
            "sum": -163608.12077951845,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 258,
          "gross_j": {
            "core": 16.176899,
            "dram": 8.684426,
            "package-0": 57.340795,
            "psys": 103.223613,
            "uncore": 23.693787
          },
          "net_j": {
            "core": 9.783602621559215,
            "dram": 7.407123382645066,
            "package-0": 46.16027016629302,
            "psys": 77.37489191726415,
            "uncore": 23.68825917944336
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.602582654,
          "token_latencies_s": [
            0.186020491,
            0.186125084,
            0.185978366,
            0.185934993,
            0.185759083,
            0.185862061,
            0.185765606,
            0.185662852,
            0.18571649,
            0.185975506,
            0.186123266,
            0.1857271,
            0.185944272,
            0.185955699
          ],
          "watts": {
            "core": 6.21570998912836,
            "dram": 3.3368492588132037,
            "package-0": 22.0322666455457,
            "psys": 39.66199222966158,
            "uncore": 9.10395178557891
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 27.678397,
            "dram": 12.288909,
            "package-0": 101.639205,
            "psys": 174.949991,
            "uncore": 48.541684000000004
          },
          "net_j": {
            "core": 18.03140653745829,
            "dram": 10.361558000137896,
            "package-0": 84.7686565297777,
            "psys": 135.94626725665168,
            "uncore": 48.53334294570939
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.95,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.854973,
            "dram": 0.628721,
            "package-0": 5.816331,
            "psys": 9.372535,
            "uncore": 1.65234
          },
          "logits": {
            "l2": 826.5092086937404,
            "sha256": "4dd5fd8e1442cc127c6d0044f29c595ddad6c04dc29eeda373180bd92ad1842c",
            "sum": 6180.617078131763,
            "top1": 78
          },
          "net_j": {
            "core": 2.377992729515987,
            "dram": 0.5334261611793915,
            "package-0": 4.982193265150395,
            "psys": 7.444057225172348,
            "uncore": 1.6519275896895405
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.194169096,
          "watts": {
            "core": 14.703539640520344,
            "dram": 3.238007556053101,
            "package-0": 29.95497800535673,
            "psys": 48.269962589721274,
            "uncore": 8.50979910829888
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 260.752459994474,
          "decode_tokens_per_s": 5.348148215410686,
          "end_to_end_ms": 3927.09622,
          "handoff_estimate_ms": 7.846010000000014,
          "handoff_plus_first_decode_ms": 194.16909600000002,
          "output_tokens_per_s_e2e": 4.074257187413656,
          "prefill_tokens_per_s": 898.0858630919743,
          "steady_tpot_ms": 186.323086,
          "steady_tpot_p95_ms": 187.25145110000003,
          "ttft_ms": 1122.3871139999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9279279279279279,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 111,
          "gross_j": {
            "core": 7.446454,
            "dram": 2.953911,
            "package-0": 37.08334,
            "psys": 59.790313,
            "uncore": 23.110963
          },
          "logits": {
            "l2": 848.6664321595354,
            "sha256": "0a12f44f7c57878debefd7f9c698d7b450eec350fae83b41138241ea19af3147",
            "sum": -208252.0210649966,
            "top1": 290
          },
          "net_j": {
            "core": 4.689287590658058,
            "dram": 2.4030627892066723,
            "package-0": 32.26163837918004,
            "psys": 48.642820177033215,
            "uncore": 23.108579077801895
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.122387114,
          "watts": {
            "core": 6.6344792336951235,
            "dram": 2.631811220170513,
            "package-0": 33.03970576412017,
            "psys": 53.270669499151076,
            "uncore": 20.590901937243736
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 760.3157061515358,
            "sha256": "779808bb2a5fa209d3cc4c2c6cffdb0cd54061013fde8ed3d471ca149f7ac757",
            "sum": -163608.12077951845,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 257,
          "gross_j": {
            "core": 17.37697,
            "dram": 8.706277,
            "package-0": 58.739534,
            "psys": 105.787143,
            "uncore": 23.778381
          },
          "net_j": {
            "core": 10.964126217284246,
            "dram": 7.425069049751833,
            "package-0": 47.52482488544726,
            "psys": 79.85938985444612,
            "uncore": 23.772836278217955
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.61054001,
          "token_latencies_s": [
            0.186082539,
            0.18625113,
            0.185821196,
            0.185915456,
            0.186131708,
            0.185764101,
            0.186256107,
            0.186390065,
            0.186693132,
            0.1866024,
            0.187131021,
            0.187003835,
            0.186996032,
            0.187475107
          ],
          "watts": {
            "core": 6.656465686576473,
            "dram": 3.3350482914069572,
            "package-0": 22.500913134826845,
            "psys": 40.523088171324375,
            "uncore": 9.108606230478728
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 27.060295,
            "dram": 12.279632000000001,
            "package-0": 100.71428,
            "psys": 171.998462,
            "uncore": 48.559934
          },
          "net_j": {
            "core": 17.37805025475524,
            "dram": 10.345237624560717,
            "package-0": 83.78207923651233,
            "psys": 132.8522017517532,
            "uncore": 48.55156246388326
          }
        },
        "first_decode": {
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 19,
          "gross_j": {
            "core": 2.759576,
            "dram": 0.627257,
            "package-0": 5.606797,
            "psys": 9.031288,
            "uncore": 1.585445
          },
          "logits": {
            "l2": 826.5092086937404,
            "sha256": "4dd5fd8e1442cc127c6d0044f29c595ddad6c04dc29eeda373180bd92ad1842c",
            "sum": 6180.617078131763,
            "top1": 78
          },
          "net_j": {
            "core": 2.2820517773761786,
            "dram": 0.5318534861781455,
            "package-0": 4.771708007763364,
            "psys": 7.100610973595076,
            "uncore": 1.5850321193735029
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.194390528,
          "watts": {
            "core": 14.196041486136608,
            "dram": 3.226787881351914,
            "package-0": 28.84295370605712,
            "psys": 46.45950650434984,
            "uncore": 8.155978669907208
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 259.8030271736732,
          "decode_tokens_per_s": 5.332785433951368,
          "end_to_end_ms": 3941.447531,
          "handoff_estimate_ms": 7.608663500000001,
          "handoff_plus_first_decode_ms": 194.39052800000002,
          "output_tokens_per_s_e2e": 4.059422299588643,
          "prefill_tokens_per_s": 893.0955975463864,
          "steady_tpot_ms": 186.7818645,
          "steady_tpot_p95_ms": 188.26474005,
          "ttft_ms": 1128.65857
        },
        "prefill": {
          "gpu_active_fraction": 0.9285714285714286,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 112,
          "gross_j": {
            "core": 5.90416,
            "dram": 2.952996,
            "package-0": 35.417817,
            "psys": 57.262793,
            "uncore": 23.223451
          },
          "logits": {
            "l2": 848.6664321595354,
            "sha256": "0a12f44f7c57878debefd7f9c698d7b450eec350fae83b41138241ea19af3147",
            "sum": -208252.0210649966,
            "top1": 290
          },
          "net_j": {
            "core": 3.1315876334752084,
            "dram": 2.399069866916245,
            "package-0": 30.569173614451174,
            "psys": 46.053012388909565,
            "uncore": 23.221053757385985
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.12865857,
          "watts": {
            "core": 5.231130261120509,
            "dram": 2.616376713464374,
            "package-0": 31.380452814884485,
            "psys": 50.73526620189487,
            "uncore": 20.576152626918873
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 760.3157061515358,
            "sha256": "779808bb2a5fa209d3cc4c2c6cffdb0cd54061013fde8ed3d471ca149f7ac757",
            "sum": -163608.12077951845,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961538461538462,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 260,
          "gross_j": {
            "core": 18.396559,
            "dram": 8.699379,
            "package-0": 59.689666,
            "psys": 105.704381,
            "uncore": 23.751038
          },
          "net_j": {
            "core": 11.964410843903853,
            "dram": 7.414314271466327,
            "package-0": 48.44119761429779,
            "psys": 79.69857838924855,
            "uncore": 23.745476587123772
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.618398433,
          "token_latencies_s": [
            0.188787433,
            0.187816799,
            0.187697824,
            0.18798329,
            0.187250099,
            0.186775363,
            0.186820822,
            0.186772544,
            0.186788366,
            0.186538855,
            0.186446257,
            0.186078308,
            0.186285922,
            0.186326829
          ],
          "watts": {
            "core": 7.025882221798596,
            "dram": 3.322404600598843,
            "package-0": 22.796250275635575,
            "psys": 40.36986108294085,
            "uncore": 9.070826540629847
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 9.113449837081134,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394373,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.047103749588131905,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 50.566337,
            "dram": 26.820427000000002,
            "package-0": 185.419019,
            "psys": 372.549583,
            "uncore": 57.687352
          },
          "net_j": {
            "core": 24.173351449699716,
            "dram": 21.54743048494852,
            "package-0": 139.26326257740027,
            "psys": 265.8401671295015,
            "uncore": 57.67187235103964
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 6.187545,
            "dram": 0.917357,
            "package-0": 8.97721,
            "psys": 15.674887,
            "uncore": 0.0
          },
          "logits": {
            "l2": 828.5850497589859,
            "sha256": "b1caee57e8f1958b838d2101b7e06beaac2d30c9baf03c086a8b6dbdc5604ef3",
            "sum": 8582.001019639429,
            "top1": 78
          },
          "net_j": {
            "core": 4.81910563361967,
            "dram": 0.6439595097376434,
            "package-0": 6.5840986199565,
            "psys": 10.142153116508794,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 241460,
          "npu_utilization": 0.43345093485848896,
          "seconds": 0.557064204,
          "watts": {
            "core": 11.107418059840011,
            "dram": 1.6467706835458413,
            "package-0": 16.115216047879464,
            "psys": 28.138384924837137,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 95.30852392037775,
          "decode_tokens_per_s": 4.340279358915426,
          "end_to_end_ms": 10744.054759,
          "handoff_estimate_ms": 350.250513,
          "handoff_plus_first_decode_ms": 557.064204,
          "output_tokens_per_s_e2e": 1.4891956862559024,
          "prefill_tokens_per_s": 138.30848685993183,
          "steady_tpot_ms": 206.81369100000003,
          "steady_tpot_p95_ms": 209.1745115,
          "ttft_ms": 7288.056018
        },
        "prefill": {
          "gpu_active_fraction": 0.9986033519553073,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 716,
          "gross_j": {
            "core": 34.430881,
            "dram": 16.89168,
            "package-0": 136.398943,
            "psys": 280.123245,
            "uncore": 57.687352
          },
          "logits": {
            "l2": 848.6442854813025,
            "sha256": "61eaa50892be757f703f9f7b089ab795b4ea224866c02f6fed0de86c01524847",
            "sum": -208360.92914666422,
            "top1": 290
          },
          "net_j": {
            "core": 16.527625526388075,
            "dram": 13.314827978446191,
            "package-0": 105.08993109466797,
            "psys": 207.73863621337873,
            "uncore": 57.67187235103964
          },
          "npu_busy_delta_us": 19331,
          "npu_utilization": 0.002652421983620379,
          "seconds": 7.288056018,
          "watts": {
            "core": 4.724288742424976,
            "dram": 2.3177209338513625,
            "package-0": 18.715408150420725,
            "psys": 38.43593467286107,
            "uncore": 7.9153277441232754
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 764.336809503711,
            "sha256": "bf80027ec35e469a6666cab0a8823ab8cf2b37944db579c08097a81568e0a72a",
            "sum": -165352.43837253097,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 286,
          "gross_j": {
            "core": 9.947911,
            "dram": 9.01139,
            "package-0": 40.042866,
            "psys": 76.751451,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.8266202896919728,
            "dram": 7.588642996764685,
            "package-0": 27.58923286277578,
            "psys": 47.95937779961395,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2871402,
          "npu_utilization": 0.9905025323446963,
          "seconds": 2.898934537,
          "token_latencies_s": [
            0.211506094,
            0.206843467,
            0.206877097,
            0.206751202,
            0.207661107,
            0.206783915,
            0.206998221,
            0.20741524,
            0.205842218,
            0.206379571,
            0.205783449,
            0.207919044,
            0.206221665,
            0.205899449
          ],
          "watts": {
            "core": 3.4315749020999706,
            "dram": 3.10851793477391,
            "package-0": 13.812959723277807,
            "psys": 26.47574480223594,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 53.838423999999996,
            "dram": 26.516778000000002,
            "package-0": 188.008918,
            "psys": 372.533104,
            "uncore": 57.885106
          },
          "net_j": {
            "core": 28.314402024978012,
            "dram": 21.4173898058191,
            "package-0": 143.37279530903015,
            "psys": 269.3369926533101,
            "uncore": 57.86969713431484
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 25,
          "gross_j": {
            "core": 3.183646,
            "dram": 0.677062,
            "package-0": 5.087938,
            "psys": 8.479104,
            "uncore": 0.0
          },
          "logits": {
            "l2": 828.5850497589859,
            "sha256": "b1caee57e8f1958b838d2101b7e06beaac2d30c9baf03c086a8b6dbdc5604ef3",
            "sum": 8582.001019639429,
            "top1": 78
          },
          "net_j": {
            "core": 2.582704359641274,
            "dram": 0.5570011918234568,
            "package-0": 4.0370180150769865,
            "psys": 6.049438311261378,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 235304,
          "npu_utilization": 0.9618720797939488,
          "seconds": 0.244631282,
          "watts": {
            "core": 13.014059256738882,
            "dram": 2.7676836521667743,
            "package-0": 20.798394867586886,
            "psys": 34.660751195343856,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 98.55329607193744,
          "decode_tokens_per_s": 4.7837933363467595,
          "end_to_end_ms": 10390.317126,
          "handoff_estimate_ms": 38.373330499999994,
          "handoff_plus_first_decode_ms": 244.631282,
          "output_tokens_per_s_e2e": 1.5398952511240225,
          "prefill_tokens_per_s": 138.9438306865149,
          "steady_tpot_ms": 206.25795150000002,
          "steady_tpot_p95_ms": 208.01400879999997,
          "ttft_ms": 7254.730167
        },
        "prefill": {
          "gpu_active_fraction": 0.9972144846796658,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 718,
          "gross_j": {
            "core": 36.730802,
            "dram": 16.842852,
            "package-0": 138.514172,
            "psys": 283.452338,
            "uncore": 57.885106
          },
          "logits": {
            "l2": 848.6442854813025,
            "sha256": "61eaa50892be757f703f9f7b089ab795b4ea224866c02f6fed0de86c01524847",
            "sum": -208360.92914666422,
            "top1": 290
          },
          "net_j": {
            "core": 18.909412146363405,
            "dram": 13.28235573054592,
            "package-0": 107.3483257705543,
            "psys": 211.39871990296155,
            "uncore": 57.86969713431484
          },
          "npu_busy_delta_us": 16959,
          "npu_utilization": 0.00233764724663949,
          "seconds": 7.254730167,
          "watts": {
            "core": 5.063014220305459,
            "dram": 2.3216372783393147,
            "package-0": 19.09294609330437,
            "psys": 39.071382597984915,
            "uncore": 7.97894679299104
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 764.336809503711,
            "sha256": "bf80027ec35e469a6666cab0a8823ab8cf2b37944db579c08097a81568e0a72a",
            "sum": -165352.43837253097,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 285,
          "gross_j": {
            "core": 13.923976,
            "dram": 8.996864,
            "package-0": 44.406808,
            "psys": 80.601662,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.822285518973332,
            "dram": 7.578032883449723,
            "package-0": 31.98745152339886,
            "psys": 51.88883443908719,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2867776,
          "npu_utilization": 0.9919820019433664,
          "seconds": 2.890955677,
          "token_latencies_s": [
            0.208843232,
            0.206246533,
            0.205858169,
            0.205696034,
            0.206504411,
            0.206781482,
            0.206689356,
            0.205705822,
            0.20597667,
            0.206553449,
            0.2061037,
            0.207567504,
            0.20626937,
            0.206126496
          ],
          "watts": {
            "core": 4.8163920708909576,
            "dram": 3.11207261722401,
            "package-0": 15.360598003384746,
            "psys": 27.88062876274945,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 51.952137,
            "dram": 26.533012,
            "package-0": 186.069957,
            "psys": 369.59732,
            "uncore": 57.989049
          },
          "net_j": {
            "core": 26.443544274883642,
            "dram": 21.43670638171136,
            "package-0": 141.46081680781074,
            "psys": 266.4635906161848,
            "uncore": 57.97365151646549
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 25,
          "gross_j": {
            "core": 2.897148,
            "dram": 0.676878,
            "package-0": 4.688159,
            "psys": 7.937113,
            "uncore": 0.0
          },
          "logits": {
            "l2": 828.5850497589859,
            "sha256": "b1caee57e8f1958b838d2101b7e06beaac2d30c9baf03c086a8b6dbdc5604ef3",
            "sum": 8582.001019639429,
            "top1": 78
          },
          "net_j": {
            "core": 2.2888436041587914,
            "dram": 0.5553462031046428,
            "package-0": 3.6243631110607097,
            "psys": 5.477678972426208,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 238245,
          "npu_utilization": 0.9621064960403268,
          "seconds": 0.247628512,
          "watts": {
            "core": 11.699573593528681,
            "dram": 2.7334412928992604,
            "package-0": 18.932226188880865,
            "psys": 32.052500481043154,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 98.61290749192062,
          "decode_tokens_per_s": 4.785200466747688,
          "end_to_end_ms": 10384.036188,
          "handoff_estimate_ms": 41.660278,
          "handoff_plus_first_decode_ms": 247.628512,
          "output_tokens_per_s_e2e": 1.5408266795612597,
          "prefill_tokens_per_s": 139.04654095136732,
          "steady_tpot_ms": 205.968234,
          "steady_tpot_p95_ms": 207.05076935,
          "ttft_ms": 7249.371276
        },
        "prefill": {
          "gpu_active_fraction": 0.99860529986053,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 717,
          "gross_j": {
            "core": 38.973655,
            "dram": 16.839495,
            "package-0": 140.998662,
            "psys": 284.459782,
            "uncore": 57.989049
          },
          "logits": {
            "l2": 848.6442854813025,
            "sha256": "61eaa50892be757f703f9f7b089ab795b4ea224866c02f6fed0de86c01524847",
            "sum": -208360.92914666422,
            "top1": 290
          },
          "net_j": {
            "core": 21.1654293693594,
            "dram": 13.281628781695265,
            "package-0": 109.85583721573857,
            "psys": 212.45938814103098,
            "uncore": 57.97365151646549
          },
          "npu_busy_delta_us": 16659,
          "npu_utilization": 0.002297992386615901,
          "seconds": 7.249371276,
          "watts": {
            "core": 5.376142773791629,
            "dram": 2.3228904078550054,
            "package-0": 19.449778005824406,
            "psys": 39.239234848095265,
            "uncore": 7.999183210822765
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 764.336809503711,
            "sha256": "bf80027ec35e469a6666cab0a8823ab8cf2b37944db579c08097a81568e0a72a",
            "sum": -165352.43837253097,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 285,
          "gross_j": {
            "core": 10.081334,
            "dram": 9.016639,
            "package-0": 40.383136,
            "psys": 77.200425,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.9892713013654477,
            "dram": 7.599731396911453,
            "package-0": 27.98061648101146,
            "psys": 48.52652350272758,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2867373,
          "npu_utilization": 0.9931890709795,
          "seconds": 2.8870364,
          "token_latencies_s": [
            0.207562021,
            0.206765845,
            0.205783042,
            0.205991599,
            0.205944869,
            0.205916718,
            0.20677548,
            0.205843808,
            0.206392005,
            0.205663847,
            0.206223223,
            0.205597092,
            0.206696886,
            0.205818322
          ],
          "watts": {
            "core": 3.49193172625049,
            "dram": 3.123146975216523,
            "package-0": 13.987747435397766,
            "psys": 26.740371198645086,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 475.13566965516657,
    "compile_target": "NPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394514,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU+GPU",
      "key": "npu_gpu_combined",
      "label": "NPU + iGPU (subgraph combination)",
      "mode": "combined",
      "prefill_device": "NPU+GPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU,GPU",
        "NPUW_SUBMODEL_DEVICE": "0:GPU,last:NPU"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 1024,
      "NPUW_LLM_MIN_RESPONSE_LEN": 1024,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04873363580554724,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 675.580913,
            "dram": 253.428733,
            "package-0": 2005.351312,
            "psys": 4124.793944,
            "uncore": 621.436885
          },
          "net_j": {
            "core": 385.5892213462486,
            "dram": 195.49193094439786,
            "package-0": 1498.2171016985849,
            "psys": 2952.3292387587444,
            "uncore": 621.1861501772278
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9058971141781681,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 797,
          "gross_j": {
            "core": 48.156676,
            "dram": 16.220113,
            "package-0": 130.372469,
            "psys": 263.966243,
            "uncore": 38.74276
          },
          "logits": {
            "l2": 899.2846808864034,
            "sha256": "80ccdd393fadb75c2187ae4c3e4330c0cb47061d068a1bf13053982ef530897e",
            "sum": -50935.694061841816,
            "top1": 15
          },
          "net_j": {
            "core": 28.31012213431361,
            "dram": 12.255013668205944,
            "package-0": 95.6650387130083,
            "psys": 183.72468919724412,
            "uncore": 38.72560012159875
          },
          "npu_busy_delta_us": 19528,
          "npu_utilization": 0.0024170908039777054,
          "seconds": 8.079133795,
          "watts": {
            "core": 5.960623653714351,
            "dram": 2.0076549555396985,
            "package-0": 16.13693649691563,
            "psys": 32.67259209933656,
            "uncore": 4.795410124780585
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 8.674305392357397,
          "decode_tokens_per_s": 0.1293517003838133,
          "end_to_end_ms": 118049.798074,
          "handoff_estimate_ms": 375.76810600000067,
          "handoff_plus_first_decode_ms": 8079.133795000001,
          "output_tokens_per_s_e2e": 0.13553602175558432,
          "prefill_tokens_per_s": 483.01657559932204,
          "steady_tpot_ms": 7703.365689,
          "steady_tpot_p95_ms": 7740.94827255,
          "ttft_ms": 2086.884904
        },
        "prefill": {
          "gpu_active_fraction": 0.13043478260869565,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 207,
          "gross_j": {
            "core": 6.064987,
            "dram": 4.264454,
            "package-0": 48.252074,
            "psys": 74.647941,
            "uncore": 1.921748
          },
          "logits": {
            "l2": 851.3160897084861,
            "sha256": "de7287cd18d1bf5585ee00e95918aa2c864ab04ccd6f05685b99b54687e84540",
            "sum": -209094.41266852617,
            "top1": 290
          },
          "net_j": {
            "core": 0.9385124657676123,
            "dram": 3.2402469317246183,
            "package-0": 39.286952965938035,
            "psys": 53.921104243303475,
            "uncore": 1.9173155085846207
          },
          "npu_busy_delta_us": 1799508,
          "npu_utilization": 0.8622938411940326,
          "seconds": 2.086884904,
          "watts": {
            "core": 2.9062393370976247,
            "dram": 2.0434543332151103,
            "package-0": 23.121578917703456,
            "psys": 35.77003257674626,
            "uncore": 0.9208691846476647
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "gpu_active_fraction": 0.9819531910893881,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10639,
          "gross_j": {
            "core": 621.35925,
            "dram": 232.944166,
            "package-0": 1826.726769,
            "psys": 3786.17976,
            "uncore": 580.772377
          },
          "net_j": {
            "core": 356.3405867461674,
            "dram": 179.9966703444673,
            "package-0": 1363.2651100196385,
            "psys": 2714.683445318197,
            "uncore": 580.5432345470444
          },
          "npu_busy_delta_us": 269916,
          "npu_utilization": 0.002501914574773859,
          "seconds": 107.883779375,
          "token_latencies_s": [
            7.685417873,
            7.709529459,
            7.669118025,
            7.697201919,
            7.726413474,
            7.722269014,
            7.673825334,
            7.690132589,
            7.709883637,
            7.731335754,
            7.691409223,
            7.734213127,
            7.7534564,
            7.689507587
          ],
          "watts": {
            "core": 5.759524310324524,
            "dram": 2.15921399259007,
            "package-0": 16.932357946511736,
            "psys": 35.09498630780611,
            "uncore": 5.38331508559092
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 662.5741710000001,
            "dram": 254.030416,
            "package-0": 1998.688121,
            "psys": 4229.072914,
            "uncore": 632.834379
          },
          "net_j": {
            "core": 373.52309784656296,
            "dram": 196.2815380450911,
            "package-0": 1493.1988537575787,
            "psys": 3060.41122113619,
            "uncore": 632.5844574619521
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9802371541501976,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 759,
          "gross_j": {
            "core": 44.132334,
            "dram": 16.673297,
            "package-0": 130.489901,
            "psys": 277.516746,
            "uncore": 42.20558
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 25.195354986903162,
            "dram": 12.889919626820383,
            "package-0": 97.37312498985148,
            "psys": 200.95269207785998,
            "uncore": 42.18920656512809
          },
          "npu_busy_delta_us": 16451,
          "npu_utilization": 0.0021340368348840193,
          "seconds": 7.708864126,
          "watts": {
            "core": 5.724881549170529,
            "dram": 2.1628733789411716,
            "package-0": 16.92725398543365,
            "psys": 35.99969352994665,
            "uncore": 5.474941484265045
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 8.702532971796606,
          "decode_tokens_per_s": 0.12980586119120155,
          "end_to_end_ms": 117666.89115899999,
          "handoff_estimate_ms": 4.4758590000002485,
          "handoff_plus_first_decode_ms": 7708.8641259999995,
          "output_tokens_per_s_e2e": 0.13597707768432196,
          "prefill_tokens_per_s": 477.7918032740029,
          "steady_tpot_ms": 7704.388266999999,
          "steady_tpot_p95_ms": 7729.5312370500005,
          "ttft_ms": 2109.70551
        },
        "prefill": {
          "gpu_active_fraction": 0.1291866028708134,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 209,
          "gross_j": {
            "core": 6.548079,
            "dram": 4.27776,
            "package-0": 48.778683,
            "psys": 77.211411,
            "uncore": 1.999385
          },
          "logits": {
            "l2": 851.3160897084861,
            "sha256": "de7287cd18d1bf5585ee00e95918aa2c864ab04ccd6f05685b99b54687e84540",
            "sum": -209094.41266852617,
            "top1": 290
          },
          "net_j": {
            "core": 1.3655451903899847,
            "dram": 3.2423529724347087,
            "package-0": 39.71552613537139,
            "psys": 56.2579211229784,
            "uncore": 1.9949040381874297
          },
          "npu_busy_delta_us": 1820083,
          "npu_utilization": 0.8627189867840843,
          "seconds": 2.10970551,
          "watts": {
            "core": 3.10378816804626,
            "dram": 2.02765740513234,
            "package-0": 23.12108622212396,
            "psys": 36.59819374505971,
            "uncore": 0.947708099790667
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "gpu_active_fraction": 0.981457078313253,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10624,
          "gross_j": {
            "core": 611.893758,
            "dram": 233.079359,
            "package-0": 1819.419537,
            "psys": 3874.344757,
            "uncore": 588.629414
          },
          "net_j": {
            "core": 346.9621976692698,
            "dram": 180.14926544583602,
            "package-0": 1356.1102026323558,
            "psys": 2803.2006079353514,
            "uncore": 588.4003468586366
          },
          "npu_busy_delta_us": 267738,
          "npu_utilization": 0.002482542113025853,
          "seconds": 107.848321523,
          "token_latencies_s": [
            7.709477718,
            7.722819643,
            7.715957626,
            7.678374276,
            7.699298816,
            7.716738099,
            7.684040405,
            7.681785127,
            7.741995626,
            7.721207771,
            7.71757316,
            7.676434782,
            7.693536295,
            7.689013931
          ],
          "watts": {
            "core": 5.673651192332243,
            "dram": 2.1611774361299907,
            "package-0": 16.870170173320556,
            "psys": 35.92401534198886,
            "uncore": 5.457937645088593
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 649.812545,
            "dram": 255.059405,
            "package-0": 1994.857687,
            "psys": 4243.751891,
            "uncore": 639.47993
          },
          "net_j": {
            "core": 357.4451683484226,
            "dram": 196.64797006550197,
            "package-0": 1483.5689054816412,
            "psys": 3061.6820595055137,
            "uncore": 639.2271410943902
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9841479524438573,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 757,
          "gross_j": {
            "core": 41.928237,
            "dram": 16.657062,
            "package-0": 127.922158,
            "psys": 271.235145,
            "uncore": 41.866775
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 22.970031583133174,
            "dram": 12.86944385031206,
            "package-0": 94.768261493564,
            "psys": 194.5852706562555,
            "uncore": 41.850383212193364
          },
          "npu_busy_delta_us": 17043,
          "npu_utilization": 0.0022083562068905744,
          "seconds": 7.717504969,
          "watts": {
            "core": 5.432874636092768,
            "dram": 2.158348075823571,
            "package-0": 16.57558479247414,
            "psys": 35.14544481532681,
            "uncore": 5.42491066324832
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 8.603820725349753,
          "decode_tokens_per_s": 0.1283144284148036,
          "end_to_end_ms": 119016.891761,
          "handoff_estimate_ms": 7.636519500000105,
          "handoff_plus_first_decode_ms": 7717.5049690000005,
          "output_tokens_per_s_e2e": 0.13443469883358988,
          "prefill_tokens_per_s": 476.2457848669109,
          "steady_tpot_ms": 7709.8684495,
          "steady_tpot_p95_ms": 8335.74722145,
          "ttft_ms": 2116.554166
        },
        "prefill": {
          "gpu_active_fraction": 0.12980769230769232,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 208,
          "gross_j": {
            "core": 6.021652,
            "dram": 4.231617,
            "package-0": 48.119384,
            "psys": 75.321462,
            "uncore": 2.060908
          },
          "logits": {
            "l2": 851.3160897084861,
            "sha256": "de7287cd18d1bf5585ee00e95918aa2c864ab04ccd6f05685b99b54687e84540",
            "sum": -209094.41266852617,
            "top1": 290
          },
          "net_j": {
            "core": 0.822294329854877,
            "dram": 3.192848770353108,
            "package-0": 39.02680575720004,
            "psys": 54.29995161324992,
            "uncore": 2.0564124918131332
          },
          "npu_busy_delta_us": 1818856,
          "npu_utilization": 0.8593477215078276,
          "seconds": 2.116554166,
          "watts": {
            "core": 2.8450261735470277,
            "dram": 1.9992953962511537,
            "package-0": 22.734775595627255,
            "psys": 35.586834114596435,
            "uncore": 0.973709075395333
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "gpu_active_fraction": 0.9822688451541032,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 10772,
          "gross_j": {
            "core": 601.862656,
            "dram": 234.170726,
            "package-0": 1818.816145,
            "psys": 3897.195284,
            "uncore": 595.552247
          },
          "net_j": {
            "core": 333.65284243543454,
            "dram": 180.5856774448368,
            "package-0": 1349.7738382308771,
            "psys": 2812.7968372360083,
            "uncore": 595.3203453903836
          },
          "npu_busy_delta_us": 270769,
          "npu_utilization": 0.0024799594724520917,
          "seconds": 109.182832626,
          "token_latencies_s": [
            7.699136824,
            7.693529823,
            7.680483029,
            7.711910971,
            7.724094863,
            7.708874295,
            7.687142327,
            7.710862604,
            7.678736704,
            7.727380883,
            7.714119771,
            7.676723884,
            8.221086883,
            8.54868785
          ],
          "watts": {
            "core": 5.512429395028141,
            "dram": 2.144757745955716,
            "package-0": 16.65844438411172,
            "psys": 35.69421300278621,
            "uncore": 5.454632680579304
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 18.58357109222561,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 1024,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 394219,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 2.4565200143075967,
      "dram": 0.4907827289910671,
      "package-0": 4.295934585025856,
      "psys": 9.931950112327097,
      "uncore": 0.002123975024632864
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04962586984038353,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 25.492611,
            "dram": 12.250333000000001,
            "package-0": 98.718008,
            "psys": 171.27605,
            "uncore": 48.225462
          },
          "net_j": {
            "core": 15.889626312733956,
            "dram": 10.331773817161723,
            "package-0": 81.92441633487049,
            "psys": 132.45024589994648,
            "uncore": 48.217158994317614
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9523809523809523,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.152702,
            "dram": 0.637327,
            "package-0": 6.115218,
            "psys": 9.851659,
            "uncore": 1.656307
          },
          "logits": {
            "l2": 826.5092086937404,
            "sha256": "4dd5fd8e1442cc127c6d0044f29c595ddad6c04dc29eeda373180bd92ad1842c",
            "sum": 6180.617078131763,
            "top1": 78
          },
          "net_j": {
            "core": 2.6514924437457474,
            "dram": 0.5371914454829739,
            "package-0": 5.23870836257903,
            "psys": 7.825219858635238,
            "uncore": 1.6558736403801348
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.204032352,
          "watts": {
            "core": 15.451971067803992,
            "dram": 3.123656585598739,
            "package-0": 29.97180564776315,
            "psys": 48.28478867900322,
            "uncore": 8.117864562968915
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 261.9473608019604,
          "decode_tokens_per_s": 5.3361626834722005,
          "end_to_end_ms": 3909.1823520000003,
          "handoff_estimate_ms": 17.914383000000004,
          "handoff_plus_first_decode_ms": 204.032352,
          "output_tokens_per_s_e2e": 4.092927512530631,
          "prefill_tokens_per_s": 917.8876628268174,
          "steady_tpot_ms": 186.117969,
          "steady_tpot_p95_ms": 186.7896282,
          "ttft_ms": 1098.173601
        },
        "prefill": {
          "gpu_active_fraction": 0.9357798165137615,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 109,
          "gross_j": {
            "core": 6.445235,
            "dram": 2.901176,
            "package-0": 35.665558,
            "psys": 57.799229,
            "uncore": 22.978151
          },
          "logits": {
            "l2": 848.6664321595354,
            "sha256": "0a12f44f7c57878debefd7f9c698d7b450eec350fae83b41138241ea19af3147",
            "sum": -208252.0210649966,
            "top1": 290
          },
          "net_j": {
            "core": 3.747549569959255,
            "dram": 2.362211363195273,
            "package-0": 30.947876047101712,
            "psys": 46.892223580193395,
            "uncore": 22.975818506698765
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.098173601,
          "watts": {
            "core": 5.8690492961504,
            "dram": 2.6418191052472766,
            "package-0": 32.47715840876418,
            "psys": 52.632142083335324,
            "uncore": 20.923969560983828
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 760.3157061515358,
            "sha256": "779808bb2a5fa209d3cc4c2c6cffdb0cd54061013fde8ed3d471ca149f7ac757",
            "sum": -163608.12077951845,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 256,
          "gross_j": {
            "core": 15.894674,
            "dram": 8.71183,
            "package-0": 56.937232,
            "psys": 103.625162,
            "uncore": 23.591004
          },
          "net_j": {
            "core": 9.490584299028953,
            "dram": 7.432371008483476,
            "package-0": 45.73783192518974,
            "psys": 77.73280246111786,
            "uncore": 23.585466847238717
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.606976399,
          "token_latencies_s": [
            0.187316019,
            0.186506187,
            0.186109384,
            0.185969027,
            0.185738382,
            0.186212148,
            0.186049341,
            0.185983335,
            0.186126554,
            0.185974281,
            0.186091623,
            0.186274996,
            0.186458708,
            0.186130159
          ],
          "watts": {
            "core": 6.096976561083167,
            "dram": 3.34173719537382,
            "package-0": 21.84033274019678,
            "psys": 39.74917534341668,
            "uncore": 9.049182036726217
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 25.780575,
            "dram": 12.320037,
            "package-0": 99.296926,
            "psys": 171.48100599999998,
            "uncore": 48.269591
          },
          "net_j": {
            "core": 16.08147532906803,
            "dram": 10.382275216040725,
            "package-0": 82.33524953195223,
            "psys": 132.26659964258357,
            "uncore": 48.26120489064917
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.95,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.924126,
            "dram": 0.630431,
            "package-0": 5.812973,
            "psys": 9.278601,
            "uncore": 1.61871
          },
          "logits": {
            "l2": 826.5092086937404,
            "sha256": "4dd5fd8e1442cc127c6d0044f29c595ddad6c04dc29eeda373180bd92ad1842c",
            "sum": 6180.617078131763,
            "top1": 78
          },
          "net_j": {
            "core": 2.4446889810453176,
            "dram": 0.5346453328076066,
            "package-0": 4.974538931043453,
            "psys": 7.340190351388662,
            "uncore": 1.6182954655169863
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.195169189,
          "watts": {
            "core": 14.982518577765878,
            "dram": 3.2301768697722055,
            "package-0": 29.784276041645082,
            "psys": 47.54132067434066,
            "uncore": 8.29388085431866
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 259.3515460192473,
          "decode_tokens_per_s": 5.3618570554846094,
          "end_to_end_ms": 3948.308833,
          "handoff_estimate_ms": 9.275330999999998,
          "handoff_plus_first_decode_ms": 195.169189,
          "output_tokens_per_s_e2e": 4.052367906550739,
          "prefill_tokens_per_s": 875.934770670514,
          "steady_tpot_ms": 185.893858,
          "steady_tpot_p95_ms": 186.3148617,
          "ttft_ms": 1150.770621
        },
        "prefill": {
          "gpu_active_fraction": 0.9122807017543859,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 114,
          "gross_j": {
            "core": 6.793196,
            "dram": 2.982414,
            "package-0": 36.264251,
            "psys": 58.664584,
            "uncore": 23.06592
          },
          "logits": {
            "l2": 848.6664321595354,
            "sha256": "0a12f44f7c57878debefd7f9c698d7b450eec350fae83b41138241ea19af3147",
            "sum": -208252.0210649966,
            "top1": 290
          },
          "net_j": {
            "core": 3.9663049376363184,
            "dram": 2.4176356541828747,
            "package-0": 31.32061568981442,
            "psys": 47.23518760149632,
            "uncore": 23.06347579194191
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.150770621,
          "watts": {
            "core": 5.903171210694299,
            "dram": 2.591666788823939,
            "package-0": 31.5130142690704,
            "psys": 50.978520766389984,
            "uncore": 20.043890223714705
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 760.3157061515358,
            "sha256": "779808bb2a5fa209d3cc4c2c6cffdb0cd54061013fde8ed3d471ca149f7ac757",
            "sum": -163608.12077951845,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 257,
          "gross_j": {
            "core": 16.063253,
            "dram": 8.707192,
            "package-0": 57.219702,
            "psys": 103.537821,
            "uncore": 23.584961
          },
          "net_j": {
            "core": 9.670481410386394,
            "dram": 7.4299942290502425,
            "package-0": 46.04009491109435,
            "psys": 77.69122168969858,
            "uncore": 23.57943363319027
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.602369023,
          "token_latencies_s": [
            0.186620168,
            0.186150466,
            0.18606484,
            0.186137962,
            0.185880094,
            0.185697566,
            0.185919071,
            0.185907622,
            0.185325541,
            0.186063856,
            0.185693032,
            0.185505621,
            0.185874779,
            0.185497544
          ],
          "watts": {
            "core": 6.172550033462337,
            "dram": 3.345871366837277,
            "package-0": 21.987543463008702,
            "psys": 39.78598733881409,
            "uncore": 9.06288108702253
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 26.783989000000002,
            "dram": 12.310272,
            "package-0": 100.133228,
            "psys": 172.820297,
            "uncore": 48.199706
          },
          "net_j": {
            "core": 17.160754385495736,
            "dram": 10.387667128751303,
            "package-0": 83.30422348976428,
            "psys": 133.9126204682395,
            "uncore": 48.19138548567146
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.781426,
            "dram": 0.628478,
            "package-0": 5.621262,
            "psys": 9.084327,
            "uncore": 1.568417
          },
          "logits": {
            "l2": 826.5092086937404,
            "sha256": "4dd5fd8e1442cc127c6d0044f29c595ddad6c04dc29eeda373180bd92ad1842c",
            "sum": 6180.617078131763,
            "top1": 78
          },
          "net_j": {
            "core": 2.3061283720301873,
            "dram": 0.5335193326252685,
            "package-0": 4.7900668514630995,
            "psys": 7.162652313040789,
            "uncore": 1.568006044548713
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.193484126,
          "watts": {
            "core": 14.375473882544762,
            "dram": 3.2482147915328206,
            "package-0": 29.052832995715626,
            "psys": 46.9512780598859,
            "uncore": 8.106179211828467
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 261.3961516494277,
          "decode_tokens_per_s": 5.382832361145902,
          "end_to_end_ms": 3917.425691,
          "handoff_estimate_ms": 8.476140500000007,
          "handoff_plus_first_decode_ms": 193.484126,
          "output_tokens_per_s_e2e": 4.0843148695223075,
          "prefill_tokens_per_s": 891.4132564637025,
          "steady_tpot_ms": 185.0079855,
          "steady_tpot_p95_ms": 186.48864155,
          "ttft_ms": 1130.788658
        },
        "prefill": {
          "gpu_active_fraction": 0.9285714285714286,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 112,
          "gross_j": {
            "core": 6.34233,
            "dram": 2.978692,
            "package-0": 35.780487,
            "psys": 57.950169,
            "uncore": 23.066225
          },
          "logits": {
            "l2": 848.6664321595354,
            "sha256": "0a12f44f7c57878debefd7f9c698d7b450eec350fae83b41138241ea19af3147",
            "sum": -208252.0210649966,
            "top1": 290
          },
          "net_j": {
            "core": 3.5645250296709716,
            "dram": 2.4237204565146135,
            "package-0": 30.922692895742827,
            "psys": 46.7192324611587,
            "uncore": 23.06382323313227
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 1.130788658,
          "watts": {
            "core": 5.608766903638328,
            "dram": 2.6341721584547413,
            "package-0": 31.64206392314204,
            "psys": 51.24756831439674,
            "uncore": 20.398351926165144
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 760.3157061515358,
            "sha256": "779808bb2a5fa209d3cc4c2c6cffdb0cd54061013fde8ed3d471ca149f7ac757",
            "sum": -163608.12077951845,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961089494163424,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 257,
          "gross_j": {
            "core": 17.660233,
            "dram": 8.703102,
            "package-0": 58.731479,
            "psys": 105.785801,
            "uncore": 23.565064
          },
          "net_j": {
            "core": 11.290100983794577,
            "dram": 7.430427339611421,
            "package-0": 47.59146374255836,
            "psys": 80.03073569404002,
            "uncore": 23.55955620799048
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.593152907,
          "token_latencies_s": [
            0.187072259,
            0.185068887,
            0.184964879,
            0.186174386,
            0.185016371,
            0.184959004,
            0.184984993,
            0.1849996,
            0.185153331,
            0.184718199,
            0.184785292,
            0.184963859,
            0.185235775,
            0.185024983
          ],
          "watts": {
            "core": 6.810332299467446,
            "dram": 3.3561854283666426,
            "package-0": 22.64867561085938,
            "psys": 40.79427816016559,
            "uncore": 9.087417844272922
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 9.288249555043876,
    "compile_target": "AUTO:GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 400412,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04650996718555689,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 33.736913,
            "dram": 15.269613,
            "package-0": 139.270579,
            "psys": 235.656014,
            "uncore": 74.074701
          },
          "net_j": {
            "core": 14.68363870787872,
            "dram": 12.775101343467147,
            "package-0": 110.34168181678035,
            "psys": 184.35023200238007,
            "uncore": 74.06483491282789
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9523809523809523,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.063713,
            "dram": 0.640867,
            "package-0": 6.03508,
            "psys": 9.682349,
            "uncore": 1.671627
          },
          "logits": {
            "l2": 653.054767718728,
            "sha256": "7147004b0f99126b2e87e0a57bfb6d03ac73ede7f2bfe0eabe33e7fdde6a26d6",
            "sum": -38815.08208623878,
            "top1": 220
          },
          "net_j": {
            "core": 2.2969966046962385,
            "dram": 0.54048619636401,
            "package-0": 4.870961989186338,
            "psys": 7.617770301958478,
            "uncore": 1.671229982108147
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.20327832,
          "watts": {
            "core": 15.0715186941726,
            "dram": 3.1526578928830182,
            "package-0": 29.688753822837572,
            "psys": 47.630996753613466,
            "uncore": 8.223341279089674
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 405.4189271720613,
          "decode_tokens_per_s": 5.386155441135341,
          "end_to_end_ms": 5051.564845,
          "handoff_estimate_ms": 19.185620000000014,
          "handoff_plus_first_decode_ms": 203.27832,
          "output_tokens_per_s_e2e": 3.1673353685317287,
          "prefill_tokens_per_s": 896.4783368403306,
          "steady_tpot_ms": 184.0927,
          "steady_tpot_p95_ms": 185.77264455,
          "ttft_ms": 2266.647075
        },
        "prefill": {
          "gpu_active_fraction": 0.9377777777777778,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 225,
          "gross_j": {
            "core": 13.091214,
            "dram": 5.904342,
            "package-0": 74.396111,
            "psys": 120.887997,
            "uncore": 48.632016
          },
          "logits": {
            "l2": 637.7699802581292,
            "sha256": "af4cff367047f08fb39636f2515d91bc86b19e3f9a6df0c7debf7735673bb18f",
            "sum": -27833.03061964945,
            "top1": 220
          },
          "net_j": {
            "core": 4.541972375168518,
            "dram": 4.785049716653374,
            "package-0": 61.4156579730092,
            "psys": 97.86699177414451,
            "uncore": 48.62758906729369
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.266647075,
          "watts": {
            "core": 5.77558550882916,
            "dram": 2.60487927967348,
            "package-0": 32.82209737040779,
            "psys": 53.33340083391677,
            "uncore": 21.455486624445054
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 838.660960244918,
            "sha256": "f7d1eadbc4a56468606230a00858318b02ca9435aea82c0d341aed77f8519197",
            "sum": -213878.9870355043,
            "top1": 42
          },
          "gpu_active_fraction": 0.99609375,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 256,
          "gross_j": {
            "core": 17.581986,
            "dram": 8.724404,
            "package-0": 58.839388,
            "psys": 105.085668,
            "uncore": 23.771058
          },
          "net_j": {
            "core": 7.8446697280139635,
            "dram": 7.449565430449764,
            "package-0": 44.05506185458481,
            "psys": 78.86546992627707,
            "uncore": 23.76601586342605
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.58163945,
          "token_latencies_s": [
            0.187462159,
            0.184862906,
            0.184317839,
            0.184206702,
            0.184031236,
            0.183943354,
            0.184059122,
            0.183863952,
            0.184126278,
            0.184272005,
            0.184053799,
            0.184319923,
            0.184058968,
            0.184018031
          ],
          "watts": {
            "core": 6.81039561895446,
            "dram": 3.3794045097970593,
            "package-0": 22.791481591281077,
            "psys": 40.70501324265091,
            "uncore": 9.207737354648806
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 35.284333,
            "dram": 15.452231000000001,
            "package-0": 141.766423,
            "psys": 240.821282,
            "uncore": 74.542717
          },
          "net_j": {
            "core": 15.749090477170023,
            "dram": 12.894618621189798,
            "package-0": 112.10574558950907,
            "psys": 188.21767808498163,
            "uncore": 74.53260134204518
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.95,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 3.246634,
            "dram": 0.634154,
            "package-0": 6.144394,
            "psys": 10.77457,
            "uncore": 1.602657
          },
          "logits": {
            "l2": 653.054767718728,
            "sha256": "7147004b0f99126b2e87e0a57bfb6d03ac73ede7f2bfe0eabe33e7fdde6a26d6",
            "sum": -38815.08208623878,
            "top1": 220
          },
          "net_j": {
            "core": 2.49492581368135,
            "dram": 0.5357381159722121,
            "package-0": 5.003063197724773,
            "psys": 8.750404717865855,
            "uncore": 1.6022677535964422
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.199299217,
          "watts": {
            "core": 16.290249650102737,
            "dram": 3.1819191743237005,
            "package-0": 30.829995684328253,
            "psys": 54.06227963253865,
            "uncore": 8.041461597914857
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 395.4165408286831,
          "decode_tokens_per_s": 5.346936994274535,
          "end_to_end_ms": 5179.3483289999995,
          "handoff_estimate_ms": 13.26480100000002,
          "handoff_plus_first_decode_ms": 199.299217,
          "output_tokens_per_s_e2e": 3.0891917252240866,
          "prefill_tokens_per_s": 855.937942878251,
          "steady_tpot_ms": 186.034416,
          "steady_tpot_p95_ms": 186.86304305000002,
          "ttft_ms": 2374.003883
        },
        "prefill": {
          "gpu_active_fraction": 0.9234042553191489,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 235,
          "gross_j": {
            "core": 13.419704,
            "dram": 6.096847,
            "package-0": 75.58904,
            "psys": 122.892019,
            "uncore": 49.05853
          },
          "logits": {
            "l2": 637.7699802581292,
            "sha256": "af4cff367047f08fb39636f2515d91bc86b19e3f9a6df0c7debf7735673bb18f",
            "sum": -27833.03061964945,
            "top1": 220
          },
          "net_j": {
            "core": 4.465538599523981,
            "dram": 4.924540884864355,
            "package-0": 61.993784598726165,
            "psys": 98.78065362709816,
            "uncore": 49.05389339134995
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.374003883,
          "watts": {
            "core": 5.652772556985746,
            "dram": 2.5681706098540533,
            "package-0": 31.840318603219405,
            "psys": 51.76571945817665,
            "uncore": 20.664890378361694
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 838.660960244918,
            "sha256": "f7d1eadbc4a56468606230a00858318b02ca9435aea82c0d341aed77f8519197",
            "sum": -213878.9870355043,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961240310077519,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 258,
          "gross_j": {
            "core": 18.617995,
            "dram": 8.72123,
            "package-0": 60.032989,
            "psys": 107.154693,
            "uncore": 23.88153
          },
          "net_j": {
            "core": 8.788626063964692,
            "dram": 7.434339620353231,
            "package-0": 45.108897793058134,
            "psys": 80.68661974001762,
            "uncore": 23.876440197098784
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.606045229,
          "token_latencies_s": [
            0.187855491,
            0.185879076,
            0.186328648,
            0.186126218,
            0.186108277,
            0.186163195,
            0.186042763,
            0.186025639,
            0.186026069,
            0.185977507,
            0.185984257,
            0.185834711,
            0.186313657,
            0.185347007
          ],
          "watts": {
            "core": 7.144156514560631,
            "dram": 3.346538234620947,
            "package-0": 23.036050307935774,
            "psys": 41.1177410919755,
            "uncore": 9.163896978550868
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 34.883395,
            "dram": 15.353293,
            "package-0": 141.180304,
            "psys": 239.899166,
            "uncore": 74.595268
          },
          "net_j": {
            "core": 15.54623908556046,
            "dram": 12.821614712415121,
            "package-0": 111.82038471851315,
            "psys": 187.82896060698423,
            "uncore": 74.58525491442806
          }
        },
        "first_decode": {
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.763421,
            "dram": 0.631468,
            "package-0": 5.656053,
            "psys": 9.213477,
            "uncore": 1.614681
          },
          "logits": {
            "l2": 653.054767718728,
            "sha256": "7147004b0f99126b2e87e0a57bfb6d03ac73ede7f2bfe0eabe33e7fdde6a26d6",
            "sum": -38815.08208623878,
            "top1": 220
          },
          "net_j": {
            "core": 2.0250140952465743,
            "dram": 0.5347935595373738,
            "package-0": 4.53491775051564,
            "psys": 7.225128798007116,
            "uncore": 1.614298641210687
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.195772669,
          "watts": {
            "core": 14.115458578132783,
            "dram": 3.225516632252687,
            "package-0": 28.89092246068321,
            "psys": 47.06212081115367,
            "uncore": 8.247734519060982
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 399.4671220941392,
          "decode_tokens_per_s": 5.359118528725976,
          "end_to_end_ms": 5126.829936,
          "handoff_estimate_ms": 10.113841000000011,
          "handoff_plus_first_decode_ms": 195.772669,
          "output_tokens_per_s_e2e": 3.1208368913604625,
          "prefill_tokens_per_s": 872.9039131132442,
          "steady_tpot_ms": 185.658828,
          "steady_tpot_p95_ms": 187.4788681,
          "ttft_ms": 2327.862173
        },
        "prefill": {
          "gpu_active_fraction": 0.9264069264069265,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 231,
          "gross_j": {
            "core": 14.699303,
            "dram": 5.99028,
            "package-0": 76.688281,
            "psys": 124.866868,
            "uncore": 49.079647
          },
          "logits": {
            "l2": 637.7699802581292,
            "sha256": "af4cff367047f08fb39636f2515d91bc86b19e3f9a6df0c7debf7735673bb18f",
            "sum": -27833.03061964945,
            "top1": 220
          },
          "net_j": {
            "core": 5.919172910148113,
            "dram": 4.840759108251377,
            "package-0": 63.357266249970934,
            "psys": 101.22413691902466,
            "uncore": 49.075100509589966
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.327862173,
          "watts": {
            "core": 6.31450743540219,
            "dram": 2.57329667945079,
            "package-0": 32.94365185769097,
            "psys": 53.64014650363924,
            "uncore": 21.083570827025934
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 838.660960244918,
            "sha256": "f7d1eadbc4a56468606230a00858318b02ca9435aea82c0d341aed77f8519197",
            "sum": -213878.9870355043,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961240310077519,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 258,
          "gross_j": {
            "core": 17.420671,
            "dram": 8.731545,
            "package-0": 58.83597,
            "psys": 105.818821,
            "uncore": 23.90094
          },
          "net_j": {
            "core": 7.602052080165773,
            "dram": 7.446062044626371,
            "package-0": 43.928200718026574,
            "psys": 79.37969488995246,
            "uncore": 23.895855763627413
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.603195094,
          "token_latencies_s": [
            0.189998441,
            0.186122175,
            0.185868193,
            0.185666476,
            0.185416776,
            0.185625292,
            0.185678032,
            0.185706653,
            0.185225968,
            0.185371103,
            0.185848897,
            0.185551696,
            0.185421231,
            0.18565118
          ],
          "watts": {
            "core": 6.692034354302605,
            "dram": 3.3541646648478203,
            "package-0": 22.601444715230397,
            "psys": 40.64959297284231,
            "uncore": 9.181386387477572
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 8.87854654993862,
    "compile_target": "NPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 400486,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 2048,
      "NPUW_LLM_MIN_RESPONSE_LEN": 2048,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.046284615993499756,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 110.79360100000001,
            "dram": 45.207525,
            "package-0": 355.566961,
            "psys": 705.92672,
            "uncore": 120.297788
          },
          "net_j": {
            "core": 40.40845321994294,
            "dram": 35.992490938545146,
            "package-0": 248.70004289779396,
            "psys": 516.3968264757608,
            "uncore": 120.26814113426224
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 55,
          "gross_j": {
            "core": 6.811567,
            "dram": 0.932981,
            "package-0": 9.547278,
            "psys": 16.596576,
            "uncore": 0.0
          },
          "logits": {
            "l2": 959.3573524085323,
            "sha256": "dd8740ccd3404cbdedc788dbda4aa438ecf1b13d968cf307f013f36b6ff4bf1f",
            "sum": -210567.55291402992,
            "top1": 82
          },
          "net_j": {
            "core": 4.6954852812920285,
            "dram": 0.6559372524234559,
            "package-0": 6.33439660579748,
            "psys": 10.898488328314416,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 246964,
          "npu_utilization": 0.4401947948058141,
          "seconds": 0.56103344,
          "watts": {
            "core": 12.141106954337696,
            "dram": 1.6629686102133234,
            "package-0": 17.017306490679058,
            "psys": 29.58215111027963,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 109.74698876480163,
          "decode_tokens_per_s": 4.308458346813925,
          "end_to_end_ms": 18661.104264,
          "handoff_estimate_ms": 352.94955649999997,
          "handoff_plus_first_decode_ms": 561.03344,
          "output_tokens_per_s_e2e": 0.8573983497250127,
          "prefill_tokens_per_s": 133.86404164804205,
          "steady_tpot_ms": 208.0838835,
          "steady_tpot_p95_ms": 210.72838845,
          "ttft_ms": 15179.580528
        },
        "prefill": {
          "gpu_active_fraction": 0.9979933110367893,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1495,
          "gross_j": {
            "core": 91.141124,
            "dram": 35.196565,
            "package-0": 302.682758,
            "psys": 606.59415,
            "uncore": 120.297788
          },
          "logits": {
            "l2": 1074.7595910452458,
            "sha256": "750c1cd564e3f8f2852142f0c33196393ada994ce8a31a875ac2c64a687452ad",
            "sum": -321836.47732180264,
            "top1": 61262
          },
          "net_j": {
            "core": 33.887437209144366,
            "dram": 27.70074108597973,
            "package-0": 215.7535513377196,
            "psys": 452.42405156887025,
            "uncore": 120.26814113426224
          },
          "npu_busy_delta_us": 19128,
          "npu_utilization": 0.0012601138723640493,
          "seconds": 15.179580528,
          "watts": {
            "core": 6.004192529028231,
            "dram": 2.318678367631899,
            "package-0": 19.940126635362315,
            "psys": 39.96119318851312,
            "uncore": 7.924974460137466
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 924.3563894551908,
            "sha256": "a6f365aaf639e7fe3d581f0c0d352c6d68d8793d9d25068091ec29dcd4ec6d3d",
            "sum": -206137.09938866645,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 289,
          "gross_j": {
            "core": 12.84091,
            "dram": 9.077979,
            "package-0": 43.336925,
            "psys": 82.735994,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.8255307295065446,
            "dram": 7.635812600141961,
            "package-0": 26.61209495427686,
            "psys": 53.07428657857611,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2904074,
          "npu_utilization": 0.9943789246543691,
          "seconds": 2.920490296,
          "token_latencies_s": [
            0.21146587,
            0.207985041,
            0.207970445,
            0.210331283,
            0.20968208,
            0.20808403,
            0.207899494,
            0.208435323,
            0.207993434,
            0.208083737,
            0.208112642,
            0.207966865,
            0.208606846,
            0.207813705
          ],
          "watts": {
            "core": 4.396833647277423,
            "dram": 3.108374991840753,
            "package-0": 14.838921074093513,
            "psys": 28.329487727905807,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 113.63166700000001,
            "dram": 44.944648,
            "package-0": 357.45581200000004,
            "psys": 698.216719,
            "uncore": 120.310238
          },
          "net_j": {
            "core": 45.199778345544594,
            "dram": 35.85579692350612,
            "package-0": 252.05224027609128,
            "psys": 511.28208964350364,
            "uncore": 120.28055127727191
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 3.204765,
            "dram": 0.718565,
            "package-0": 5.130786,
            "psys": 8.59269,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1160.1109936426578,
            "sha256": "ee26dafc622fa1586c41834e36486e7569534a9d8a43aa7dc2efbf4ca2a106ad",
            "sum": -332855.33728590235,
            "top1": 82
          },
          "net_j": {
            "core": 2.123772498588445,
            "dram": 0.5770382285877615,
            "package-0": 3.489497450341238,
            "psys": 5.681843036472221,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 233955,
          "npu_utilization": 0.8163066552120303,
          "seconds": 0.286601853,
          "watts": {
            "core": 11.18194096253802,
            "dram": 2.5071889538690457,
            "package-0": 17.902138267054397,
            "psys": 29.981278592780065,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 111.27063598008455,
          "decode_tokens_per_s": 4.679330396542526,
          "end_to_end_ms": 18405.574678,
          "handoff_estimate_ms": 78.151252,
          "handoff_plus_first_decode_ms": 286.601853,
          "output_tokens_per_s_e2e": 0.8693018435944105,
          "prefill_tokens_per_s": 133.68431760569518,
          "steady_tpot_ms": 208.450601,
          "steady_tpot_p95_ms": 209.06685405,
          "ttft_ms": 15199.987824999998
        },
        "prefill": {
          "gpu_active_fraction": 0.9973333333333333,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1500,
          "gross_j": {
            "core": 100.406664,
            "dram": 35.132295,
            "package-0": 311.894892,
            "psys": 613.700517,
            "uncore": 120.310238
          },
          "logits": {
            "l2": 1099.655919480962,
            "sha256": "53b6512b8a4023eb3df87a5f0f2dea980cdd9c64787beb63d76b70823888e6a9",
            "sum": -334396.9358225651,
            "top1": 81842
          },
          "net_j": {
            "core": 43.07600584695615,
            "dram": 27.626393765029327,
            "package-0": 224.84881846283736,
            "psys": 459.3231536196577,
            "uncore": 120.28055127727191
          },
          "npu_busy_delta_us": 19386,
          "npu_utilization": 0.0012753957584173263,
          "seconds": 15.199987825,
          "watts": {
            "core": 6.605706870031655,
            "dram": 2.311337048718985,
            "package-0": 20.51941722525689,
            "psys": 40.37506635305479,
            "uncore": 7.915153576775974
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 934.8850718863815,
            "sha256": "cff6f4d6adb86e0011bd27d99cdcbdb1ddbe055e012bb2cc498d9ba52a16da5a",
            "sum": -179314.85746893287,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 286,
          "gross_j": {
            "core": 10.020238,
            "dram": 9.093788,
            "package-0": 40.430134,
            "psys": 75.923512,
            "uncore": 0.0
          },
          "net_j": {
            "core": 0.0,
            "dram": 7.65236492988903,
            "package-0": 23.713924362912692,
            "psys": 46.27709298737369,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2899695,
          "npu_utilization": 0.9933915384971145,
          "seconds": 2.918985,
          "token_latencies_s": [
            0.209580447,
            0.208711221,
            0.208790304,
            0.20868243,
            0.208524595,
            0.208054801,
            0.208055263,
            0.208466085,
            0.208239135,
            0.208435117,
            0.208543522,
            0.208278713,
            0.208287669,
            0.208276503
          ],
          "watts": {
            "core": 3.4327816004535823,
            "dram": 3.1153938783515502,
            "package-0": 13.850750860316172,
            "psys": 26.010243971791564,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 110.363609,
            "dram": 45.050301,
            "package-0": 355.26935499999996,
            "psys": 704.231841,
            "uncore": 121.057064
          },
          "net_j": {
            "core": 40.9080154736985,
            "dram": 35.95696696163944,
            "package-0": 249.81379428554956,
            "psys": 517.205008480759,
            "uncore": 121.02736083404473
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 29,
          "gross_j": {
            "core": 3.170829,
            "dram": 0.717406,
            "package-0": 5.09685,
            "psys": 8.829385,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1051.2152043955496,
            "sha256": "88a3b6b82b708f14d483f7b8106973fef372581f06ddb88a53c39b31214058f2",
            "sum": -237789.0732052317,
            "top1": 82
          },
          "net_j": {
            "core": 2.0810303063112947,
            "dram": 0.5747262955565027,
            "package-0": 3.4421908649470683,
            "psys": 5.894825126330188,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 235245,
          "npu_utilization": 0.8141750807811229,
          "seconds": 0.288936625,
          "watts": {
            "core": 10.974133168475959,
            "dram": 2.4829181831829037,
            "package-0": 17.640027462769734,
            "psys": 30.558206319465384,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 111.21578019057145,
          "decode_tokens_per_s": 4.678368371722701,
          "end_to_end_ms": 18414.652997,
          "handoff_estimate_ms": 80.7656015,
          "handoff_plus_first_decode_ms": 288.936625,
          "output_tokens_per_s_e2e": 0.8688732827388395,
          "prefill_tokens_per_s": 133.61031197249645,
          "steady_tpot_ms": 208.1710235,
          "steady_tpot_p95_ms": 209.62374945000002,
          "ttft_ms": 15208.406971
        },
        "prefill": {
          "gpu_active_fraction": 0.9973368841544608,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1502,
          "gross_j": {
            "core": 95.607849,
            "dram": 35.240938,
            "package-0": 308.008855,
            "psys": 613.770646,
            "uncore": 121.057064
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 38.245435875384864,
            "dram": 27.730879309175656,
            "package-0": 220.9145673714661,
            "psys": 459.3077742905487,
            "uncore": 121.02736083404473
          },
          "npu_busy_delta_us": 16935,
          "npu_utilization": 0.0011135288549479466,
          "seconds": 15.208406971,
          "watts": {
            "core": 6.286513057041995,
            "dram": 2.3172011419209673,
            "package-0": 20.252538979744795,
            "psys": 40.357326521466874,
            "uncore": 7.95987799582405
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 965.9590353683316,
            "sha256": "dbeca01a9b1e044fd4348fb2ab0d9cd6a49005015a4454119b4d452579f76fc8",
            "sum": -210379.17602261715,
            "top1": 271
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 288,
          "gross_j": {
            "core": 11.584931,
            "dram": 9.091957,
            "package-0": 42.16365,
            "psys": 81.63181,
            "uncore": 0.0
          },
          "net_j": {
            "core": 0.5815492920023395,
            "dram": 7.651361356907277,
            "package-0": 25.45703604913639,
            "psys": 52.00240906388014,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2900361,
          "npu_utilization": 0.9941903998958115,
          "seconds": 2.917309401,
          "token_latencies_s": [
            0.208650638,
            0.208241794,
            0.207759065,
            0.207900049,
            0.207908723,
            0.211129791,
            0.208812804,
            0.208250814,
            0.208139036,
            0.20862718,
            0.208203011,
            0.207872886,
            0.207792791,
            0.207987724
          ],
          "watts": {
            "core": 3.9711012469328413,
            "dram": 3.116555616926832,
            "package-0": 14.45292363763236,
            "psys": 27.9818828856542,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 10.806858167983592,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 2048,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 400326,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.05028646066784859,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 36.02719,
            "dram": 15.214623,
            "package-0": 141.730838,
            "psys": 236.40179999999998,
            "uncore": 74.245109
          },
          "net_j": {
            "core": 16.975593615401422,
            "dram": 12.720331020138454,
            "package-0": 112.80448841114459,
            "psys": 185.10053619469977,
            "uncore": 74.23524378167498
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.95,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 3.109001,
            "dram": 0.636168,
            "package-0": 6.015365,
            "psys": 9.582312,
            "uncore": 1.611079
          },
          "logits": {
            "l2": 653.054767718728,
            "sha256": "7147004b0f99126b2e87e0a57bfb6d03ac73ede7f2bfe0eabe33e7fdde6a26d6",
            "sum": -38815.08208623878,
            "top1": 220
          },
          "net_j": {
            "core": 2.348080885976656,
            "dram": 0.53654606284527,
            "package-0": 4.860047577612,
            "psys": 7.533341261997814,
            "uncore": 1.6106849835143888
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.201741561,
          "watts": {
            "core": 15.41081066583003,
            "dram": 3.1533809733929834,
            "package-0": 29.81718278664454,
            "psys": 47.497957052092005,
            "uncore": 7.985855725583485
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 405.4546331283506,
          "decode_tokens_per_s": 5.39349008766773,
          "end_to_end_ms": 5051.119984,
          "handoff_estimate_ms": 17.535054500000008,
          "handoff_plus_first_decode_ms": 201.74156100000002,
          "output_tokens_per_s_e2e": 3.1676143213152392,
          "prefill_tokens_per_s": 895.1583477055146,
          "steady_tpot_ms": 184.20650650000002,
          "steady_tpot_p95_ms": 184.4843545,
          "ttft_ms": 2269.9894440000003
        },
        "prefill": {
          "gpu_active_fraction": 0.9422222222222222,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 225,
          "gross_j": {
            "core": 14.193506,
            "dram": 5.882676,
            "package-0": 75.624501,
            "psys": 121.158076,
            "uncore": 48.775754
          },
          "logits": {
            "l2": 637.7699802581292,
            "sha256": "af4cff367047f08fb39636f2515d91bc86b19e3f9a6df0c7debf7735673bb18f",
            "sum": -27833.03061964945,
            "top1": 220
          },
          "net_j": {
            "core": 5.631657772007612,
            "dram": 4.761733222462611,
            "package-0": 62.624907161772235,
            "psys": 98.10312429210748,
            "uncore": 48.77132053939493
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.269989444,
          "watts": {
            "core": 6.252674891293458,
            "dram": 2.5914992757120503,
            "package-0": 33.3149130714636,
            "psys": 53.37384996227321,
            "uncore": 21.48721621984776
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 838.660960244918,
            "sha256": "f7d1eadbc4a56468606230a00858318b02ca9435aea82c0d341aed77f8519197",
            "sum": -213878.9870355043,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 254,
          "gross_j": {
            "core": 18.724683,
            "dram": 8.695779,
            "package-0": 60.090972,
            "psys": 105.661412,
            "uncore": 23.858276
          },
          "net_j": {
            "core": 8.995854957417153,
            "dram": 7.422051734830573,
            "package-0": 45.31953367176036,
            "psys": 79.46407064059447,
            "uncore": 23.85323825876566
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.579388979,
          "token_latencies_s": [
            0.184430114,
            0.184394104,
            0.184389861,
            0.184090122,
            0.184158812,
            0.184108165,
            0.184100188,
            0.184089887,
            0.184023195,
            0.184024977,
            0.184305586,
            0.184254201,
            0.184456505,
            0.184536075
          ],
          "watts": {
            "core": 7.259348300099874,
            "dram": 3.371255390635675,
            "package-0": 23.296591746816176,
            "psys": 40.963737094419834,
            "uncore": 9.249584376083357
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 38.191735,
            "dram": 15.395040999999999,
            "package-0": 144.725277,
            "psys": 241.94530500000002,
            "uncore": 74.830131
          },
          "net_j": {
            "core": 18.822873067621305,
            "dram": 12.85921166571738,
            "package-0": 115.31721795424016,
            "psys": 189.78972309796757,
            "uncore": 74.82010149654985
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.95,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 3.275138,
            "dram": 0.632811,
            "package-0": 6.159408,
            "psys": 9.686011,
            "uncore": 1.606503
          },
          "logits": {
            "l2": 653.054767718728,
            "sha256": "7147004b0f99126b2e87e0a57bfb6d03ac73ede7f2bfe0eabe33e7fdde6a26d6",
            "sum": -38815.08208623878,
            "top1": 220
          },
          "net_j": {
            "core": 2.5273248428573662,
            "dram": 0.5349050648415319,
            "package-0": 5.023991084064256,
            "psys": 7.672334073552924,
            "uncore": 1.6061157705042344
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.198266534,
          "watts": {
            "core": 16.51886444940829,
            "dram": 3.1917186790585648,
            "package-0": 31.066301890363405,
            "psys": 48.85348426981631,
            "uncore": 8.102744157518787
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 398.8132111011527,
          "decode_tokens_per_s": 5.377329538516507,
          "end_to_end_ms": 5135.236103,
          "handoff_estimate_ms": 13.366095500000007,
          "handoff_plus_first_decode_ms": 198.266534,
          "output_tokens_per_s_e2e": 3.1157282117277556,
          "prefill_tokens_per_s": 866.2484289655399,
          "steady_tpot_ms": 184.90043849999998,
          "steady_tpot_p95_ms": 186.01432140000003,
          "ttft_ms": 2345.7474
        },
        "prefill": {
          "gpu_active_fraction": 0.9310344827586207,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 232,
          "gross_j": {
            "core": 13.889674,
            "dram": 6.032516,
            "package-0": 76.030445,
            "psys": 123.071036,
            "uncore": 49.209773
          },
          "logits": {
            "l2": 637.7699802581292,
            "sha256": "af4cff367047f08fb39636f2515d91bc86b19e3f9a6df0c7debf7735673bb18f",
            "sum": -27833.03061964945,
            "top1": 220
          },
          "net_j": {
            "core": 5.0420851827509665,
            "dram": 4.874163209899049,
            "package-0": 62.597006563676246,
            "psys": 99.24665515591252,
            "uncore": 49.20519157839258
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.3457474,
          "watts": {
            "core": 5.921214705385584,
            "dram": 2.571681844345218,
            "package-0": 32.41203421988233,
            "psys": 52.46559625303219,
            "uncore": 20.97829160974451
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 838.660960244918,
            "sha256": "f7d1eadbc4a56468606230a00858318b02ca9435aea82c0d341aed77f8519197",
            "sum": -213878.9870355043,
            "top1": 42
          },
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 2450,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 256,
          "gross_j": {
            "core": 21.026923,
            "dram": 8.729714,
            "package-0": 62.535424,
            "psys": 109.188258,
            "uncore": 24.013855
          },
          "net_j": {
            "core": 11.253463042012973,
            "dram": 7.450143390976799,
            "package-0": 47.69622030649966,
            "psys": 82.87073386850211,
            "uncore": 24.008794147653035
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.591222169,
          "token_latencies_s": [
            0.186240784,
            0.185870879,
            0.185493526,
            0.18589238,
            0.185625148,
            0.184926683,
            0.184856874,
            0.184846511,
            0.184874194,
            0.184971079,
            0.184532287,
            0.18444911,
            0.184151947,
            0.184458015
          ],
          "watts": {
            "core": 8.11467393709227,
            "dram": 3.368956203152953,
            "package-0": 24.133563207408635,
            "psys": 42.13774461575317,
            "uncore": 9.26738559406019
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.497951,
            "dram": 15.420187,
            "package-0": 144.016417,
            "psys": 244.079149,
            "uncore": 74.733269
          },
          "net_j": {
            "core": 18.05887825143846,
            "dram": 12.875165455670038,
            "package-0": 114.50175572664597,
            "psys": 191.73450663656214,
            "uncore": 74.72320314027732
          }
        },
        "first_decode": {
          "gpu_active_fraction": 1.0,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 20,
          "gross_j": {
            "core": 2.898857,
            "dram": 0.632078,
            "package-0": 5.825852,
            "psys": 9.335608,
            "uncore": 1.641537
          },
          "logits": {
            "l2": 653.054767718728,
            "sha256": "7147004b0f99126b2e87e0a57bfb6d03ac73ede7f2bfe0eabe33e7fdde6a26d6",
            "sum": -38815.08208623878,
            "top1": 220
          },
          "net_j": {
            "core": 2.1585769879917653,
            "dram": 0.5351583267333204,
            "package-0": 4.701872781213739,
            "psys": 7.342215980813286,
            "uncore": 1.6411536712860864
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.196269283,
          "watts": {
            "core": 14.769794619364866,
            "dram": 3.220463183737213,
            "package-0": 29.682953496090374,
            "psys": 47.56530343059337,
            "uncore": 8.363697950636524
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 397.37276168167176,
          "decode_tokens_per_s": 5.357202494286338,
          "end_to_end_ms": 5153.850987,
          "handoff_estimate_ms": 10.983376999999988,
          "handoff_plus_first_decode_ms": 196.269283,
          "output_tokens_per_s_e2e": 3.1044747006380606,
          "prefill_tokens_per_s": 863.2547699795491,
          "steady_tpot_ms": 185.285906,
          "steady_tpot_p95_ms": 189.0231269,
          "ttft_ms": 2353.882157
        },
        "prefill": {
          "gpu_active_fraction": 0.9227467811158798,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 233,
          "gross_j": {
            "core": 15.558615,
            "dram": 6.047897,
            "package-0": 77.693039,
            "psys": 126.280256,
            "uncore": 49.119625
          },
          "logits": {
            "l2": 637.7699802581292,
            "sha256": "af4cff367047f08fb39636f2515d91bc86b19e3f9a6df0c7debf7735673bb18f",
            "sum": -27833.03061964945,
            "top1": 220
          },
          "net_j": {
            "core": 6.680343857337652,
            "dram": 4.88552718808854,
            "package-0": 64.21301508846526,
            "psys": 102.3732551991166,
            "uncore": 49.11502769059857
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.353882157,
          "watts": {
            "core": 6.609768018221143,
            "dram": 2.569328707477857,
            "package-0": 33.006341786888356,
            "psys": 53.647654205826065,
            "uncore": 20.867495364594838
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 838.660960244918,
            "sha256": "f7d1eadbc4a56468606230a00858318b02ca9435aea82c0d341aed77f8519197",
            "sum": -213878.9870355043,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961089494163424,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 257,
          "gross_j": {
            "core": 19.040479,
            "dram": 8.740212,
            "package-0": 60.497526,
            "psys": 108.463285,
            "uncore": 23.972107
          },
          "net_j": {
            "core": 9.219957406109042,
            "dram": 7.454479940848177,
            "package-0": 45.586867856966975,
            "psys": 82.01903545663225,
            "uncore": 23.96702177839266
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.603699547,
          "token_latencies_s": [
            0.195117783,
            0.185645203,
            0.185324044,
            0.184661946,
            0.184984062,
            0.185136193,
            0.185247768,
            0.185070948,
            0.18500667,
            0.185058867,
            0.185633582,
            0.185741389,
            0.185335556,
            0.18569932
          ],
          "watts": {
            "core": 7.312855671822261,
            "dram": 3.3568435382917086,
            "package-0": 23.235217776838212,
            "psys": 41.65737368774831,
            "uncore": 9.206940573316464
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.165887873619795,
    "compile_target": "AUTO:GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 400833,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.046541888266801834,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 52.242725,
            "dram": 22.619206000000002,
            "package-0": 238.720031,
            "psys": 398.251788,
            "uncore": 138.88233499999998
          },
          "net_j": {
            "core": 21.227217440000445,
            "dram": 18.558562986337286,
            "package-0": 191.62867990033,
            "psys": 314.7346536207989,
            "uncore": 138.8662746789978
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9523809523809523,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.197868,
            "dram": 0.653136,
            "package-0": 6.228011,
            "psys": 9.966955,
            "uncore": 1.698054
          },
          "logits": {
            "l2": 728.3785859021272,
            "sha256": "c0ed3cbfd820a639c3f89b51903efbb9d58551be551f5f0d06ab29ea3c24d06a",
            "sum": -48270.276146711665,
            "top1": 78
          },
          "net_j": {
            "core": 2.395075716024573,
            "dram": 0.5480320331201751,
            "package-0": 5.009118378892325,
            "psys": 7.8052328057183455,
            "uncore": 1.6976383014416647
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.212843064,
          "watts": {
            "core": 15.024534696606324,
            "dram": 3.068627127074247,
            "package-0": 29.261047472986952,
            "psys": 46.82771809749929,
            "uncore": 7.977962579978645
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 498.109405927577,
          "decode_tokens_per_s": 5.255019366956979,
          "end_to_end_ms": 8223.093062,
          "handoff_estimate_ms": 24.365791499999983,
          "handoff_plus_first_decode_ms": 212.843064,
          "output_tokens_per_s_e2e": 1.9457398669045975,
          "prefill_tokens_per_s": 759.963452791308,
          "steady_tpot_ms": 188.47727250000003,
          "steady_tpot_p95_ms": 189.7162846,
          "ttft_ms": 5368.679224
        },
        "prefill": {
          "gpu_active_fraction": 0.9510357815442562,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 531,
          "gross_j": {
            "core": 30.181746,
            "dram": 13.134427,
            "package-0": 171.24413,
            "psys": 279.519731,
            "uncore": 112.567095
          },
          "logits": {
            "l2": 680.59636228507,
            "sha256": "8abfd8de30eb1ada57815fb449ec5279eb380a6c63325e4eb5ea55f3d3b40f0e",
            "sum": 21399.900736933458,
            "top1": 49
          },
          "net_j": {
            "core": 9.93239337760783,
            "dram": 10.483321191131829,
            "package-0": 140.49920757571238,
            "psys": 224.99320420418383,
            "uncore": 112.55660956452637
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 5.368679224,
          "watts": {
            "core": 5.621819583683885,
            "dram": 2.4464912973910247,
            "package-0": 31.896882427706768,
            "psys": 52.06489703285725,
            "uncore": 20.96737210462921
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 811.5720666048574,
            "sha256": "8248a9d79387f416fa16b5c5c188e01f3867aea1d96b374479ac88edb673436c",
            "sum": -176094.05456782482,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961685823754789,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 261,
          "gross_j": {
            "core": 18.863111,
            "dram": 8.831643,
            "package-0": 61.24789,
            "psys": 108.765102,
            "uncore": 24.617186
          },
          "net_j": {
            "core": 8.89974834636804,
            "dram": 7.527209762085283,
            "package-0": 46.12035394572532,
            "psys": 81.93621661089674,
            "uncore": 24.61202681302977
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.641570774,
          "token_latencies_s": [
            0.191071571,
            0.188986515,
            0.188773333,
            0.188380882,
            0.188574069,
            0.188533096,
            0.188342663,
            0.188324014,
            0.188261792,
            0.188255816,
            0.188421449,
            0.18863195,
            0.188624163,
            0.188356265
          ],
          "watts": {
            "core": 7.140869056268564,
            "dram": 3.343330069717072,
            "package-0": 23.18616279481899,
            "psys": 41.17440390790756,
            "uncore": 9.319146866060837
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 52.845262,
            "dram": 22.969302,
            "package-0": 240.888543,
            "psys": 405.645873,
            "uncore": 139.672922
          },
          "net_j": {
            "core": 21.220425088392957,
            "dram": 18.828883765189023,
            "package-0": 192.87203720549314,
            "psys": 320.48796452606746,
            "uncore": 139.6565461586097
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9047619047619048,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 2.450555,
            "dram": 0.643919,
            "package-0": 5.419114,
            "psys": 9.094276,
            "uncore": 1.659236
          },
          "logits": {
            "l2": 728.3785859021272,
            "sha256": "c0ed3cbfd820a639c3f89b51903efbb9d58551be551f5f0d06ab29ea3c24d06a",
            "sum": -48270.276146711665,
            "top1": 78
          },
          "net_j": {
            "core": 1.674865436743472,
            "dram": 0.542363402377894,
            "package-0": 4.241371881899924,
            "psys": 7.005534767385454,
            "uncore": 1.6588343356596744
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.205657362,
          "watts": {
            "core": 11.915717366830757,
            "dram": 3.1310282001964023,
            "package-0": 26.35020670935184,
            "psys": 44.22052248243853,
            "uncore": 8.067963061784289
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 488.51211749912517,
          "decode_tokens_per_s": 5.252636397745488,
          "end_to_end_ms": 8384.643601,
          "handoff_estimate_ms": 16.53028850000002,
          "handoff_plus_first_decode_ms": 205.657362,
          "output_tokens_per_s_e2e": 1.9082504589809577,
          "prefill_tokens_per_s": 737.9359947373357,
          "steady_tpot_ms": 189.1270735,
          "steady_tpot_p95_ms": 190.72192264999998,
          "ttft_ms": 5528.934798
        },
        "prefill": {
          "gpu_active_fraction": 0.9396709323583181,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 547,
          "gross_j": {
            "core": 32.6511,
            "dram": 13.490811,
            "package-0": 175.292642,
            "psys": 288.125178,
            "uncore": 113.326737
          },
          "logits": {
            "l2": 680.59636228507,
            "sha256": "8abfd8de30eb1ada57815fb449ec5279eb380a6c63325e4eb5ea55f3d3b40f0e",
            "sum": 21399.900736933458,
            "top1": 49
          },
          "net_j": {
            "core": 11.7973023191057,
            "dram": 10.760569436245293,
            "package-0": 143.62998078277724,
            "psys": 231.97102930961674,
            "uncore": 113.31593857330257
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 5.528934798,
          "watts": {
            "core": 5.905495577884368,
            "dram": 2.4400379987986254,
            "package-0": 31.70459562362884,
            "psys": 52.11224015595635,
            "uncore": 20.497029019223387
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 811.5720666048574,
            "sha256": "8248a9d79387f416fa16b5c5c188e01f3867aea1d96b374479ac88edb673436c",
            "sum": -176094.05456782482,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961977186311787,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 263,
          "gross_j": {
            "core": 17.743607,
            "dram": 8.834572,
            "package-0": 60.176787,
            "psys": 108.426419,
            "uncore": 24.686949
          },
          "net_j": {
            "core": 7.748257332543785,
            "dram": 7.525950926565837,
            "package-0": 45.00068454081597,
            "psys": 81.51140044906529,
            "uncore": 24.68177324964744
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.650051441,
          "token_latencies_s": [
            0.192755135,
            0.189123625,
            0.188554996,
            0.188807943,
            0.18857951,
            0.188662776,
            0.188701032,
            0.188920403,
            0.189178376,
            0.189627116,
            0.189189275,
            0.189130522,
            0.189423868,
            0.189375524
          ],
          "watts": {
            "core": 6.695570782318259,
            "dram": 3.3337360412393595,
            "package-0": 22.707780712849942,
            "psys": 40.91483558488403,
            "uncore": 9.31564897875505
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 60.733304000000004,
            "dram": 22.895631,
            "package-0": 248.924534,
            "psys": 407.235835,
            "uncore": 139.902901
          },
          "net_j": {
            "core": 29.099991449114118,
            "dram": 18.75410310914495,
            "package-0": 200.8951595041088,
            "psys": 322.0551037138587,
            "uncore": 139.88652076978946
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9523809523809523,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.448172,
            "dram": 0.646787,
            "package-0": 6.452315,
            "psys": 10.056493,
            "uncore": 1.687007
          },
          "logits": {
            "l2": 728.3785859021272,
            "sha256": "c0ed3cbfd820a639c3f89b51903efbb9d58551be551f5f0d06ab29ea3c24d06a",
            "sum": -48270.276146711665,
            "top1": 78
          },
          "net_j": {
            "core": 2.659387330722775,
            "dram": 0.5435169519316696,
            "package-0": 5.254690368801754,
            "psys": 7.932489867281941,
            "uncore": 1.6865985548063918
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.209129247,
          "watts": {
            "core": 16.488234187540492,
            "dram": 3.0927620563755966,
            "package-0": 30.853240723426886,
            "psys": 48.087453784022856,
            "uncore": 8.066815255161321
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 488.3812285040317,
          "decode_tokens_per_s": 5.254028638369279,
          "end_to_end_ms": 8386.890734,
          "handoff_estimate_ms": 20.805042999999994,
          "handoff_plus_first_decode_ms": 209.129247,
          "output_tokens_per_s_e2e": 1.9077391738438738,
          "prefill_tokens_per_s": 737.5352939832277,
          "steady_tpot_ms": 188.324204,
          "steady_tpot_p95_ms": 191.93740460000004,
          "ttft_ms": 5531.938652
        },
        "prefill": {
          "gpu_active_fraction": 0.9397810218978102,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 548,
          "gross_j": {
            "core": 37.900171,
            "dram": 13.417629,
            "package-0": 180.586513,
            "psys": 288.725396,
            "uncore": 113.474563
          },
          "logits": {
            "l2": 680.59636228507,
            "sha256": "8abfd8de30eb1ada57815fb449ec5279eb380a6c63325e4eb5ea55f3d3b40f0e",
            "sum": 21399.900736933458,
            "top1": 49
          },
          "net_j": {
            "core": 17.035043512213853,
            "dram": 10.685904104040828,
            "package-0": 148.90664955214461,
            "psys": 232.54073892593652,
            "uncore": 113.46375870654914
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 5.531938652,
          "watts": {
            "core": 6.851155333455784,
            "dram": 2.4254840561453137,
            "package-0": 32.64434484187768,
            "psys": 52.1924435831578,
            "uncore": 20.512621368093942
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 811.5720666048574,
            "sha256": "8248a9d79387f416fa16b5c5c188e01f3867aea1d96b374479ac88edb673436c",
            "sum": -176094.05456782482,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961832061068703,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 262,
          "gross_j": {
            "core": 19.384961,
            "dram": 8.831215,
            "package-0": 61.885706,
            "psys": 108.453946,
            "uncore": 24.741331
          },
          "net_j": {
            "core": 9.40556060617749,
            "dram": 7.524682053172452,
            "package-0": 46.733819583162415,
            "psys": 81.58187492064027,
            "uncore": 24.736163508433915
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.645822835,
          "token_latencies_s": [
            0.197799427,
            0.188780931,
            0.188150261,
            0.188389867,
            0.188369775,
            0.188309607,
            0.188420428,
            0.188143914,
            0.188330012,
            0.188234266,
            0.188356951,
            0.188318396,
            0.187895585,
            0.188292806
          ],
          "watts": {
            "core": 7.326628504209731,
            "dram": 3.3377952911953,
            "package-0": 23.389965942296357,
            "psys": 40.99063042518491,
            "uncore": 9.351091340172818
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 12.078973207622766,
    "compile_target": "NPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 400902,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 4096,
      "NPUW_LLM_MIN_RESPONSE_LEN": 4096,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04649261571466923,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 217.701347,
            "dram": 82.94302499999999,
            "package-0": 688.033577,
            "psys": 1369.036046,
            "uncore": 248.241003
          },
          "net_j": {
            "core": 85.4405387400066,
            "dram": 65.62704440534506,
            "package-0": 487.21983418283855,
            "psys": 1012.890215557502,
            "uncore": 248.17957328269597
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 61,
          "gross_j": {
            "core": 7.139874,
            "dram": 1.003538,
            "package-0": 9.827428,
            "psys": 17.088335,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1172.8521223979067,
            "sha256": "f34cfb4134e278790a81609bd0bc150951ab37a8c8a15deb20eea3ad3a0caeae",
            "sum": -332944.55895423517,
            "top1": 82
          },
          "net_j": {
            "core": 4.830995527933675,
            "dram": 0.7012527248402955,
            "package-0": 6.321820149984649,
            "psys": 10.871093085007123,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 257571,
          "npu_utilization": 0.42076497078061054,
          "seconds": 0.612149342,
          "watts": {
            "core": 11.6636145955377,
            "dram": 1.639367930579267,
            "package-0": 16.053971352631134,
            "psys": 27.915304040300676,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 116.80796638475361,
          "decode_tokens_per_s": 4.151343714530478,
          "end_to_end_ms": 35066.101455,
          "handoff_estimate_ms": 397.9884635000001,
          "handoff_plus_first_decode_ms": 612.149342,
          "output_tokens_per_s_e2e": 0.4562811186904438,
          "prefill_tokens_per_s": 129.71812497716138,
          "steady_tpot_ms": 214.1608785,
          "steady_tpot_p95_ms": 215.79169374999998,
          "ttft_ms": 31452.813557999998
        },
        "prefill": {
          "gpu_active_fraction": 0.9977426636568849,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3101,
          "gross_j": {
            "core": 197.483625,
            "dram": 72.595334,
            "package-0": 633.880884,
            "psys": 1267.20867,
            "uncore": 248.241003
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 78.85126010564711,
            "dram": 57.06362997192164,
            "package-0": 453.7594290724311,
            "psys": 947.7608859701123,
            "uncore": 248.17957328269597
          },
          "npu_busy_delta_us": 19029,
          "npu_utilization": 0.0006050015196545108,
          "seconds": 31.452813558,
          "watts": {
            "core": 6.2787268501698215,
            "dram": 2.3080712275908755,
            "package-0": 20.153392091016066,
            "psys": 40.289199173333934,
            "uncore": 7.892489571473013
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 826.7915182092064,
            "sha256": "ff3906e0aa910a2693afc8b2817af7b14250e5373ac7ea4a033206fe356d605d",
            "sum": -21699.585335694253,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 296,
          "gross_j": {
            "core": 13.077848,
            "dram": 9.344153,
            "package-0": 44.325265,
            "psys": 84.739041,
            "uncore": 0.0
          },
          "net_j": {
            "core": 1.7582831064258162,
            "dram": 7.8621617085831215,
            "package-0": 27.138584960422794,
            "psys": 54.25823650238256,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2986510,
          "npu_utilization": 0.9951256648995335,
          "seconds": 3.001138555,
          "token_latencies_s": [
            0.216395339,
            0.215466654,
            0.214270064,
            0.213999575,
            0.214034035,
            0.214142197,
            0.21418248,
            0.214227899,
            0.21417956,
            0.213963466,
            0.214273048,
            0.214073312,
            0.213934869,
            0.213939565
          ],
          "watts": {
            "core": 4.357628866621921,
            "dram": 3.113536022664572,
            "package-0": 14.769483043744378,
            "psys": 28.23563106035936,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 212.465277,
            "dram": 82.555391,
            "package-0": 682.927145,
            "psys": 1356.3416200000001,
            "uncore": 248.884861
          },
          "net_j": {
            "core": 81.11902542490313,
            "dram": 65.35914690186827,
            "package-0": 483.5019885171326,
            "psys": 1002.6584657978357,
            "uncore": 248.82336498299983
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 33,
          "gross_j": {
            "core": 3.821524,
            "dram": 0.793577,
            "package-0": 5.950363,
            "psys": 10.358677,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1057.9006125110022,
            "sha256": "d8d60d86e793894196a424af650e125bbb024e6c0c26858765388ce2d8633963",
            "sum": -231576.80951821245,
            "top1": 82
          },
          "net_j": {
            "core": 2.543966654530762,
            "dram": 0.6263153713003077,
            "package-0": 4.010626840948888,
            "psys": 6.918529251578265,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 241875,
          "npu_utilization": 0.7140921623597257,
          "seconds": 0.338716783,
          "watts": {
            "core": 11.282357981062898,
            "dram": 2.3428924689568746,
            "package-0": 17.56736984597542,
            "psys": 30.582119103321784,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 117.62129379398894,
          "decode_tokens_per_s": 4.495235132572752,
          "end_to_end_ms": 34823.62647,
          "handoff_estimate_ms": 124.68929550000001,
          "handoff_plus_first_decode_ms": 338.716783,
          "output_tokens_per_s_e2e": 0.4594581788827693,
          "prefill_tokens_per_s": 129.57827409433824,
          "steady_tpot_ms": 214.0274875,
          "steady_tpot_p95_ms": 214.77195360000002,
          "ttft_ms": 31486.759864
        },
        "prefill": {
          "gpu_active_fraction": 0.9980707395498393,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3110,
          "gross_j": {
            "core": 196.680649,
            "dram": 72.418149,
            "package-0": 633.6188,
            "psys": 1261.939335,
            "uncore": 248.884861
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 77.92024689356174,
            "dram": 56.86968195717047,
            "package-0": 453.30294408458735,
            "psys": 942.1467782454814,
            "uncore": 248.82336498299983
          },
          "npu_busy_delta_us": 16748,
          "npu_utilization": 0.0005319061114048962,
          "seconds": 31.486759864,
          "watts": {
            "core": 6.246455648327042,
            "dram": 2.2999555785604477,
            "package-0": 20.12334081807002,
            "psys": 40.078412019866896,
            "uncore": 7.9044291021052135
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 996.0453905212725,
            "sha256": "00b38d5947b99dea850265253a33360c423c4e3867809feaf8734e71936f6b65",
            "sum": -159259.00955941156,
            "top1": 42
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 295,
          "gross_j": {
            "core": 11.963104,
            "dram": 9.343665,
            "package-0": 43.357982,
            "psys": 84.043608,
            "uncore": 0.0
          },
          "net_j": {
            "core": 0.654811876810621,
            "dram": 7.863149573397492,
            "package-0": 26.188417591596323,
            "psys": 53.593158300776,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2983086,
          "npu_utilization": 0.994975627006883,
          "seconds": 2.998149823,
          "token_latencies_s": [
            0.21544929,
            0.21400437,
            0.213782058,
            0.214050605,
            0.21398043,
            0.21436247,
            0.213995224,
            0.21408853,
            0.214407234,
            0.214144598,
            0.213863075,
            0.21415792,
            0.213957722,
            0.213877195
          ],
          "watts": {
            "core": 3.990162168756968,
            "dram": 3.116477011362484,
            "package-0": 14.461579493921109,
            "psys": 28.031823945310556,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 210.614574,
            "dram": 82.62387299999999,
            "package-0": 680.556912,
            "psys": 1357.55524,
            "uncore": 248.698704
          },
          "net_j": {
            "core": 79.7855965848623,
            "dram": 65.40994807070463,
            "package-0": 480.9267105956118,
            "psys": 1003.508435915967,
            "uncore": 248.63716569851644
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 34,
          "gross_j": {
            "core": 3.490104,
            "dram": 0.793943,
            "package-0": 5.607468,
            "psys": 9.764013,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1037.37398392076,
            "sha256": "f24b0a92963b9d23a311db8388f6d69f6cb2c438ee75a7260dd6294ce7f2dab9",
            "sum": -205472.12200375088,
            "top1": 82
          },
          "net_j": {
            "core": 2.1852290002479373,
            "dram": 0.6231048622975575,
            "package-0": 3.626255000804863,
            "psys": 6.250305526680981,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 243407,
          "npu_utilization": 0.7035708547015638,
          "seconds": 0.34595947,
          "watts": {
            "core": 10.088187497801405,
            "dram": 2.2949017698518266,
            "package-0": 16.208453550931846,
            "psys": 28.222996757394732,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 117.50048216936753,
          "decode_tokens_per_s": 4.476247218038092,
          "end_to_end_ms": 34859.43142,
          "handoff_estimate_ms": 131.443986,
          "handoff_plus_first_decode_ms": 345.95947,
          "output_tokens_per_s_e2e": 0.4589862584740919,
          "prefill_tokens_per_s": 129.4892376690935,
          "steady_tpot_ms": 214.51548400000001,
          "steady_tpot_p95_ms": 215.484725,
          "ttft_ms": 31508.410069
        },
        "prefill": {
          "gpu_active_fraction": 0.9980713596914176,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 3111,
          "gross_j": {
            "core": 196.442429,
            "dram": 72.496946,
            "package-0": 633.068814,
            "psys": 1269.667526,
            "uncore": 248.698704
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 77.60036758461436,
            "dram": 56.93778787620923,
            "package-0": 452.62897342366387,
            "psys": 949.6550808085283,
            "uncore": 248.63716569851644
          },
          "npu_busy_delta_us": 19616,
          "npu_utilization": 0.0006225639426757202,
          "seconds": 31.508410069,
          "watts": {
            "core": 6.23460303359682,
            "dram": 2.3008760467836855,
            "package-0": 20.0920583619944,
            "psys": 40.296147067388226,
            "uncore": 7.893089605453808
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 868.6798604910134,
            "sha256": "eb0b0977dcabeb495be6d6188eaaa16de8d04d91612c28c0136baa3d5e134f7f",
            "sum": -136328.59814876132,
            "top1": 198
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 296,
          "gross_j": {
            "core": 10.682041,
            "dram": 9.332984,
            "package-0": 41.88063,
            "psys": 78.123701,
            "uncore": 0.0
          },
          "net_j": {
            "core": 0.0,
            "dram": 7.849055332197847,
            "package-0": 24.671482171143076,
            "psys": 47.6030495807577,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2985494,
          "npu_utilization": 0.9934883600488492,
          "seconds": 3.005061881,
          "token_latencies_s": [
            0.215764355,
            0.21479991,
            0.214122807,
            0.214506947,
            0.214260313,
            0.21434419,
            0.214436798,
            0.215334155,
            0.215167918,
            0.214775767,
            0.214524021,
            0.214526227,
            0.214384702,
            0.214059003
          ],
          "watts": {
            "core": 3.5546825399965862,
            "dram": 3.1057543470267057,
            "package-0": 13.936694703292865,
            "psys": 25.997368471494713,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 6.344236670061946,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 4096,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 400748,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04840262979269028,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 54.762800000000006,
            "dram": 22.649478000000002,
            "package-0": 241.683646,
            "psys": 398.893754,
            "uncore": 138.984142
          },
          "net_j": {
            "core": 23.604732300709408,
            "dram": 18.570170586519904,
            "package-0": 194.37584351519035,
            "psys": 314.99274022474987,
            "uncore": 138.96800785909997
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9523809523809523,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.131828,
            "dram": 0.650511,
            "package-0": 6.156295,
            "psys": 9.829443,
            "uncore": 1.696468
          },
          "logits": {
            "l2": 728.3785859021272,
            "sha256": "c0ed3cbfd820a639c3f89b51903efbb9d58551be551f5f0d06ab29ea3c24d06a",
            "sum": -48270.276146711665,
            "top1": 78
          },
          "net_j": {
            "core": 2.335063040042409,
            "dram": 0.5461961484076409,
            "package-0": 4.946553763275324,
            "psys": 7.683950907030431,
            "uncore": 1.6960554224854951
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.211245049,
          "watts": {
            "core": 14.825568763980831,
            "dram": 3.079414183098795,
            "package-0": 29.142907865263155,
            "psys": 46.53099822472052,
            "uncore": 8.030805966960202
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 495.83036388374614,
          "decode_tokens_per_s": 5.226943087414251,
          "end_to_end_ms": 8260.889809,
          "handoff_estimate_ms": 21.54377900000001,
          "handoff_plus_first_decode_ms": 211.245049,
          "output_tokens_per_s_e2e": 1.9368373589208834,
          "prefill_tokens_per_s": 756.7967553502504,
          "steady_tpot_ms": 189.70126999999997,
          "steady_tpot_p95_ms": 190.83485255,
          "ttft_ms": 5391.143621
        },
        "prefill": {
          "gpu_active_fraction": 0.9512195121951219,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 533,
          "gross_j": {
            "core": 32.187662,
            "dram": 13.156704,
            "package-0": 173.71214,
            "psys": 280.83888,
            "uncore": 112.717485
          },
          "logits": {
            "l2": 680.59636228507,
            "sha256": "8abfd8de30eb1ada57815fb449ec5279eb380a6c63325e4eb5ea55f3d3b40f0e",
            "sum": 21399.900736933458,
            "top1": 49
          },
          "net_j": {
            "core": 11.853579134409678,
            "dram": 10.494505054273663,
            "package-0": 142.83857026521824,
            "psys": 226.08419549651217,
            "uncore": 112.70695568986459
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 5.391143621,
          "watts": {
            "core": 5.97047013821337,
            "dram": 2.440429141741093,
            "package-0": 32.22176076395795,
            "psys": 52.092635578479985,
            "uncore": 20.907898754715813
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 811.5720666048574,
            "sha256": "8248a9d79387f416fa16b5c5c188e01f3867aea1d96b374479ac88edb673436c",
            "sum": -176094.05456782482,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961977186311787,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 263,
          "gross_j": {
            "core": 19.44331,
            "dram": 8.842263,
            "package-0": 61.815211,
            "psys": 108.225431,
            "uncore": 24.570189
          },
          "net_j": {
            "core": 9.416090126257322,
            "dram": 7.529469383838599,
            "package-0": 46.590719486696784,
            "psys": 81.22459382120726,
            "uncore": 24.5649967467499
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.658501139,
          "token_latencies_s": [
            0.192278823,
            0.189756143,
            0.189596364,
            0.189646397,
            0.189956999,
            0.189913206,
            0.189472156,
            0.189847858,
            0.189479864,
            0.189825747,
            0.189606108,
            0.18951307,
            0.19005733,
            0.189516134
          ],
          "watts": {
            "core": 7.313636136832213,
            "dram": 3.326033180984844,
            "package-0": 23.251903146918302,
            "psys": 40.70919113493748,
            "uncore": 9.24212092278513
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 57.066809,
            "dram": 22.827029,
            "package-0": 244.579024,
            "psys": 403.873782,
            "uncore": 139.383982
          },
          "net_j": {
            "core": 25.56360958713523,
            "dram": 18.702535910328493,
            "package-0": 196.74720240416625,
            "psys": 319.0434133965132,
            "uncore": 139.3676691444329
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9523809523809523,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.328116,
            "dram": 0.64514,
            "package-0": 6.30046,
            "psys": 9.924474,
            "uncore": 1.657466
          },
          "logits": {
            "l2": 728.3785859021272,
            "sha256": "c0ed3cbfd820a639c3f89b51903efbb9d58551be551f5f0d06ab29ea3c24d06a",
            "sum": -48270.276146711665,
            "top1": 78
          },
          "net_j": {
            "core": 2.5566841769166784,
            "dram": 0.5441418384557939,
            "package-0": 5.129182478263525,
            "psys": 7.847197814588357,
            "uncore": 1.6570665403849654
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.204528514,
          "watts": {
            "core": 16.272136998951648,
            "dram": 3.154279016567832,
            "package-0": 30.804800156128845,
            "psys": 48.523669418533984,
            "uncore": 8.10383827459872
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 490.39831931942734,
          "decode_tokens_per_s": 5.265535644353759,
          "end_to_end_ms": 8352.394041,
          "handoff_estimate_ms": 15.836338999999978,
          "handoff_plus_first_decode_ms": 204.528514,
          "output_tokens_per_s_e2e": 1.915618434841513,
          "prefill_tokens_per_s": 741.3220337162451,
          "steady_tpot_ms": 188.69217500000002,
          "steady_tpot_p95_ms": 189.93227535000003,
          "ttft_ms": 5503.681011000001
        },
        "prefill": {
          "gpu_active_fraction": 0.9357798165137615,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 545,
          "gross_j": {
            "core": 34.04752,
            "dram": 13.337063,
            "package-0": 176.172156,
            "psys": 284.549686,
            "uncore": 113.102738
          },
          "logits": {
            "l2": 680.59636228507,
            "sha256": "8abfd8de30eb1ada57815fb449ec5279eb380a6c63325e4eb5ea55f3d3b40f0e",
            "sum": 21399.900736933458,
            "top1": 49
          },
          "net_j": {
            "core": 13.288973463135267,
            "dram": 10.619292000907143,
            "package-0": 144.65411614863638,
            "psys": 228.65202521587787,
            "uncore": 113.09198889585345
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 5.503681011,
          "watts": {
            "core": 6.186317835635913,
            "dram": 2.423298692882766,
            "package-0": 32.009877688748915,
            "psys": 51.70170390167621,
            "uncore": 20.550380331626382
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 811.5720666048574,
            "sha256": "8248a9d79387f416fa16b5c5c188e01f3867aea1d96b374479ac88edb673436c",
            "sum": -176094.05456782482,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961832061068703,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 262,
          "gross_j": {
            "core": 19.691173,
            "dram": 8.844826,
            "package-0": 62.106408,
            "psys": 109.399622,
            "uncore": 24.623778
          },
          "net_j": {
            "core": 9.717951947083284,
            "dram": 7.539102070965555,
            "package-0": 46.96390377726633,
            "psys": 82.54419036604693,
            "uncore": 24.6186137081945
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.644184516,
          "token_latencies_s": [
            0.191668738,
            0.188904542,
            0.188871058,
            0.188997257,
            0.18868876,
            0.188934233,
            0.188523841,
            0.188769778,
            0.188291469,
            0.18869559,
            0.188642575,
            0.188393052,
            0.188472906,
            0.18830558
          ],
          "watts": {
            "core": 7.44697387071455,
            "dram": 3.34501088955019,
            "package-0": 23.48792515204336,
            "psys": 41.37367167004467,
            "uncore": 9.312428028755615
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 55.82255,
            "dram": 22.86365,
            "package-0": 243.433154,
            "psys": 404.320864,
            "uncore": 139.325571
          },
          "net_j": {
            "core": 24.320423361532924,
            "dram": 18.73929736116118,
            "package-0": 195.60296121503495,
            "psys": 319.49338411414436,
            "uncore": 139.30925869993249
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9047619047619048,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 3.177299,
            "dram": 0.645933,
            "package-0": 6.207503,
            "psys": 9.969884,
            "uncore": 1.664913
          },
          "logits": {
            "l2": 728.3785859021272,
            "sha256": "c0ed3cbfd820a639c3f89b51903efbb9d58551be551f5f0d06ab29ea3c24d06a",
            "sum": -48270.276146711665,
            "top1": 78
          },
          "net_j": {
            "core": 2.4035227811003246,
            "dram": 0.5446279031436917,
            "package-0": 5.032665943850265,
            "psys": 7.886294933199791,
            "uncore": 1.664512326419684
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.20515008,
          "watts": {
            "core": 15.487681018696167,
            "dram": 3.14858760961731,
            "package-0": 30.25835037451606,
            "psys": 48.59800200906575,
            "uncore": 8.115585428969855
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 490.41501935900476,
          "decode_tokens_per_s": 5.250463004729424,
          "end_to_end_ms": 8352.109618,
          "handoff_estimate_ms": 15.860385000000004,
          "handoff_plus_first_decode_ms": 205.15008,
          "output_tokens_per_s_e2e": 1.9156836693711123,
          "prefill_tokens_per_s": 742.4636238054565,
          "steady_tpot_ms": 189.289695,
          "steady_tpot_p95_ms": 190.30641555,
          "ttft_ms": 5495.218714
        },
        "prefill": {
          "gpu_active_fraction": 0.9393382352941176,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 544,
          "gross_j": {
            "core": 34.733676,
            "dram": 13.364956,
            "package-0": 176.808447,
            "psys": 286.563536,
            "uncore": 113.034745
          },
          "logits": {
            "l2": 680.59636228507,
            "sha256": "8abfd8de30eb1ada57815fb449ec5279eb380a6c63325e4eb5ea55f3d3b40f0e",
            "sum": 21399.900736933458,
            "top1": 49
          },
          "net_j": {
            "core": 14.007047189786249,
            "dram": 10.65136376514263,
            "package-0": 145.33886835370024,
            "psys": 230.75182180439108,
            "uncore": 113.02401242335777
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 5.495218714,
          "watts": {
            "core": 6.320708566432503,
            "dram": 2.432106290137372,
            "package-0": 32.17496085270465,
            "psys": 52.14779445810426,
            "uncore": 20.569653526623945
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 811.5720666048574,
            "sha256": "8248a9d79387f416fa16b5c5c188e01f3867aea1d96b374479ac88edb673436c",
            "sum": -176094.05456782482,
            "top1": 42
          },
          "gpu_active_fraction": 0.9961977186311787,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 263,
          "gross_j": {
            "core": 17.911575,
            "dram": 8.852761,
            "package-0": 60.417204,
            "psys": 107.787444,
            "uncore": 24.625913
          },
          "net_j": {
            "core": 7.909853390646353,
            "dram": 7.543305692874861,
            "package-0": 45.231426917484455,
            "psys": 80.85526737655351,
            "uncore": 24.620733950155017
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.651740824,
          "token_latencies_s": [
            0.191527748,
            0.189013559,
            0.189173941,
            0.189315444,
            0.189422637,
            0.18911767,
            0.188783522,
            0.188939389,
            0.189095198,
            0.189648775,
            0.189266767,
            0.189312623,
            0.189542561,
            0.189549645
          ],
          "watts": {
            "core": 6.7546476781925495,
            "dram": 3.3384714372825144,
            "package-0": 22.78397777534838,
            "psys": 40.64780502847513,
            "uncore": 9.286696790696617
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.547535102814436,
    "compile_target": "AUTO:GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 401325,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "GPU",
      "key": "gpu_gpu",
      "label": "iGPU -> iGPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04683802369982004,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 88.304279,
            "dram": 41.468399999999995,
            "package-0": 464.982573,
            "psys": 772.664464,
            "uncore": 287.207883
          },
          "net_j": {
            "core": 26.533168689207045,
            "dram": 33.381141487642005,
            "package-0": 371.1944885124335,
            "psys": 606.3300549757737,
            "uncore": 287.1758969408219
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9230769230769231,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 26,
          "gross_j": {
            "core": 3.228386,
            "dram": 0.704893,
            "package-0": 6.784772,
            "psys": 11.126131,
            "uncore": 1.946162
          },
          "logits": {
            "l2": 842.3149968162743,
            "sha256": "d805104637e3082ed7ef6942f1d19e2c0a800de34252afd57ab996e71250bf5f",
            "sum": -27488.31816223578,
            "top1": 78
          },
          "net_j": {
            "core": 2.233616222199197,
            "dram": 0.5746547654453855,
            "package-0": 5.2743968181934004,
            "psys": 8.44746062685719,
            "uncore": 1.9456468924550452
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.263741757,
          "watts": {
            "core": 12.240708626203622,
            "dram": 2.672663623758296,
            "package-0": 25.725058015746818,
            "psys": 42.18570137151244,
            "uncore": 7.379043887995331
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 500.2052243362155,
          "decode_tokens_per_s": 4.952966834959828,
          "end_to_end_ms": 16377.277968000002,
          "handoff_estimate_ms": 66.8084195,
          "handoff_plus_first_decode_ms": 263.741757,
          "output_tokens_per_s_e2e": 0.9769633287816709,
          "prefill_tokens_per_s": 612.4899666880626,
          "steady_tpot_ms": 196.93333750000002,
          "steady_tpot_p95_ms": 200.4805697,
          "ttft_ms": 13348.790094
        },
        "prefill": {
          "gpu_active_fraction": 0.9587786259541985,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1310,
          "gross_j": {
            "core": 67.152294,
            "dram": 31.687907,
            "package-0": 395.736108,
            "psys": 649.431382,
            "uncore": 259.146736
          },
          "logits": {
            "l2": 1025.1392851150267,
            "sha256": "4ff86f4268bae4a3871f14f73062d317d9c93b8e3d5513020b94dee3023132a8",
            "sum": -317470.8328210273,
            "top1": 8
          },
          "net_j": {
            "core": 16.80390356397988,
            "dram": 25.0961451377224,
            "package-0": 319.29132544278286,
            "psys": 513.8555484693997,
            "uncore": 259.1206648060728
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 13.348790094,
          "watts": {
            "core": 5.030590302725903,
            "dram": 2.37384113293107,
            "package-0": 29.645840949875677,
            "psys": 48.650954687788946,
            "uncore": 19.413499963302364
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 727.8313868092002,
            "sha256": "474801b73ad22489ab599f54a4d0eb647b9530fa6cbcb6caa0057490e9977584",
            "sum": -118284.7835280532,
            "top1": 42
          },
          "gpu_active_fraction": 0.9963235294117647,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 272,
          "gross_j": {
            "core": 17.923599,
            "dram": 9.0756,
            "package-0": 62.461693,
            "psys": 112.106951,
            "uncore": 26.114985
          },
          "net_j": {
            "core": 7.49564890302797,
            "dram": 7.710341584474218,
            "package-0": 46.62876625145723,
            "psys": 84.02704587951683,
            "uncore": 26.109585242294013
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.764746117,
          "token_latencies_s": [
            0.205273177,
            0.196863468,
            0.197899935,
            0.197039062,
            0.196485823,
            0.197023833,
            0.197219045,
            0.196131999,
            0.196767983,
            0.196370331,
            0.197443275,
            0.197003207,
            0.196450385,
            0.196733348
          ],
          "watts": {
            "core": 6.482909548110236,
            "dram": 3.2826160580154276,
            "package-0": 22.592198472016154,
            "psys": 40.54873259814764,
            "uncore": 9.445708175308742
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 119.203919,
            "dram": 43.560619,
            "package-0": 502.080136,
            "psys": 836.679561,
            "uncore": 288.51897499999995
          },
          "net_j": {
            "core": 51.37893799813362,
            "dram": 34.68076963031782,
            "package-0": 399.10036088238843,
            "psys": 654.0435670011412,
            "uncore": 288.48385415070385
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9565217391304348,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 23,
          "gross_j": {
            "core": 2.824029,
            "dram": 0.680601,
            "package-0": 6.076828,
            "psys": 9.977208,
            "uncore": 1.801876
          },
          "logits": {
            "l2": 842.3149968162743,
            "sha256": "d805104637e3082ed7ef6942f1d19e2c0a800de34252afd57ab996e71250bf5f",
            "sum": -27488.31816223578,
            "top1": 78
          },
          "net_j": {
            "core": 1.9392499431477224,
            "dram": 0.5647630796018874,
            "package-0": 4.733453524142605,
            "psys": 7.594715588934793,
            "uncore": 1.8014178473922577
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.234580089,
          "watts": {
            "core": 12.038656017391144,
            "dram": 2.9013587764475615,
            "package-0": 25.90513127480312,
            "psys": 42.53220314875062,
            "uncore": 7.681282787815807
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 455.5582859604053,
          "decode_tokens_per_s": 4.994483062377571,
          "end_to_end_ms": 17982.331246,
          "handoff_estimate_ms": 37.42046249999997,
          "handoff_plus_first_decode_ms": 234.580089,
          "output_tokens_per_s_e2e": 0.8897622772664167,
          "prefill_tokens_per_s": 545.8301947938577,
          "steady_tpot_ms": 197.15962650000003,
          "steady_tpot_p95_ms": 200.97298634999999,
          "ttft_ms": 14979.017427
        },
        "prefill": {
          "gpu_active_fraction": 0.8695064232589588,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1479,
          "gross_j": {
            "core": 99.275564,
            "dram": 33.811498,
            "package-0": 434.594835,
            "psys": 715.445557,
            "uncore": 260.794925
          },
          "logits": {
            "l2": 1025.1392851150267,
            "sha256": "4ff86f4268bae4a3871f14f73062d317d9c93b8e3d5513020b94dee3023132a8",
            "sum": -317470.8328210273,
            "top1": 8
          },
          "net_j": {
            "core": 42.77835245859381,
            "dram": 26.414714088170353,
            "package-0": 348.8141970497138,
            "psys": 563.312460379677,
            "uncore": 260.76566984912654
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 14.979017427,
          "watts": {
            "core": 6.627641932043798,
            "dram": 2.257257404551386,
            "package-0": 29.013574296043842,
            "psys": 47.76318343220524,
            "uncore": 17.410683061888395
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 727.8313868092002,
            "sha256": "474801b73ad22489ab599f54a4d0eb647b9530fa6cbcb6caa0057490e9977584",
            "sum": -118284.7835280532,
            "top1": 42
          },
          "gpu_active_fraction": 0.9963369963369964,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 273,
          "gross_j": {
            "core": 17.104326,
            "dram": 9.06852,
            "package-0": 61.408473,
            "psys": 111.256796,
            "uncore": 25.922174
          },
          "net_j": {
            "core": 6.661335596392092,
            "dram": 7.701292462545577,
            "package-0": 45.55271030853201,
            "psys": 83.13639103252953,
            "uncore": 25.916766454185044
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.76873373,
          "token_latencies_s": [
            0.206184388,
            0.197772983,
            0.198166847,
            0.197165967,
            0.197465865,
            0.1964097,
            0.196840054,
            0.196136329,
            0.19656939,
            0.197153286,
            0.197059042,
            0.197635826,
            0.197367107,
            0.196766749
          ],
          "watts": {
            "core": 6.177670974521627,
            "dram": 3.2753312107047576,
            "package-0": 22.1792627924535,
            "psys": 40.183277573607626,
            "uncore": 9.362465490677573
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 98.41021,
            "dram": 41.596268,
            "package-0": 475.068426,
            "psys": 784.449713,
            "uncore": 287.320493
          },
          "net_j": {
            "core": 36.06931586162302,
            "dram": 33.43441168477512,
            "package-0": 380.41522943184714,
            "psys": 616.5810162520157,
            "uncore": 287.28821189773237
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9130434782608695,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 23,
          "gross_j": {
            "core": 3.272147,
            "dram": 0.681761,
            "package-0": 6.495162,
            "psys": 10.508335,
            "uncore": 1.788021
          },
          "logits": {
            "l2": 842.3149968162743,
            "sha256": "d805104637e3082ed7ef6942f1d19e2c0a800de34252afd57ab996e71250bf5f",
            "sum": -27488.31816223578,
            "top1": 78
          },
          "net_j": {
            "core": 2.3839941330503915,
            "dram": 0.565481370293257,
            "package-0": 5.146665013233194,
            "psys": 8.116757748045972,
            "uncore": 1.787561100379959
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.235474582,
          "watts": {
            "core": 13.89596691162191,
            "dram": 2.8952636594976524,
            "package-0": 27.583282853008736,
            "psys": 44.626196639771514,
            "uncore": 7.593265416646966
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 495.6334444277145,
          "decode_tokens_per_s": 5.00980860741616,
          "end_to_end_ms": 16528.343864,
          "handoff_estimate_ms": 38.78060449999998,
          "handoff_plus_first_decode_ms": 235.474582,
          "output_tokens_per_s_e2e": 0.9680340711478799,
          "prefill_tokens_per_s": 604.0984634963498,
          "steady_tpot_ms": 196.69397750000002,
          "steady_tpot_p95_ms": 198.88595655000003,
          "ttft_ms": 13534.217505999999
        },
        "prefill": {
          "gpu_active_fraction": 0.9514200298953662,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1338,
          "gross_j": {
            "core": 78.102462,
            "dram": 31.843119,
            "package-0": 407.319148,
            "psys": 664.025033,
            "uncore": 259.633918
          },
          "logits": {
            "l2": 1025.1392851150267,
            "sha256": "4ff86f4268bae4a3871f14f73062d317d9c93b8e3d5513020b94dee3023132a8",
            "sum": -317470.8328210273,
            "top1": 8
          },
          "net_j": {
            "core": 27.05468445315004,
            "dram": 25.15979128538408,
            "package-0": 329.8124745829619,
            "psys": 526.5659219765113,
            "uncore": 259.60748465235076
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 13.534217506,
          "watts": {
            "core": 5.770740862216494,
            "dram": 2.3527861131153895,
            "package-0": 30.095507761673474,
            "psys": 49.062683727790244,
            "uncore": 19.183518950016794
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 727.8313868092002,
            "sha256": "474801b73ad22489ab599f54a4d0eb647b9530fa6cbcb6caa0057490e9977584",
            "sum": -118284.7835280532,
            "top1": 42
          },
          "gpu_active_fraction": 0.9963235294117647,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 272,
          "gross_j": {
            "core": 17.035601,
            "dram": 9.071388,
            "package-0": 61.254116,
            "psys": 109.916345,
            "uncore": 25.898554
          },
          "net_j": {
            "core": 6.630637275422586,
            "dram": 7.709139029097786,
            "package-0": 45.456089835652044,
            "psys": 81.89833652745844,
            "uncore": 25.893166145001672
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.758651776,
          "token_latencies_s": [
            0.200611,
            0.196887108,
            0.196252684,
            0.197060868,
            0.196389091,
            0.196500847,
            0.197099661,
            0.19613736,
            0.197644193,
            0.197957087,
            0.19644485,
            0.196186573,
            0.196951478,
            0.196468223
          ],
          "watts": {
            "core": 6.175335773876231,
            "dram": 3.288341094341876,
            "package-0": 22.204366833431028,
            "psys": 39.84422606588531,
            "uncore": 9.388120032152981
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 15.612135223113,
    "compile_target": "NPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "NPU"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 401394,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": "NPU",
      "key": "gpu_npu",
      "label": "iGPU -> NPU",
      "mode": "staged",
      "prefill_device": "GPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "GPU",
        "NPUW_ONLINE_PIPELINE": "NONE"
      },
      "NPUW_LLM": "YES",
      "NPUW_LLM_GENERATE_HINT": "FAST_COMPILE",
      "NPUW_LLM_MAX_GENERATION_TOKEN_LEN": 1,
      "NPUW_LLM_MAX_PROMPT_LEN": 8192,
      "NPUW_LLM_MIN_RESPONSE_LEN": 8192,
      "NPU_USE_NPUW": "YES",
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.0465422049164772,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 388.932598,
            "dram": 158.13967000000002,
            "package-0": 1322.617158,
            "psys": 2648.284846,
            "uncore": 517.074421
          },
          "net_j": {
            "core": 133.16714723749644,
            "dram": 124.65409222037042,
            "package-0": 934.2843016789285,
            "psys": 1959.5713861423321,
            "uncore": 516.9494647944705
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 67,
          "gross_j": {
            "core": 8.604042,
            "dram": 1.106992,
            "package-0": 11.576936,
            "psys": 19.474925,
            "uncore": 0.0
          },
          "logits": {
            "l2": 963.664214914989,
            "sha256": "37bca8e9d81de23e1b20414b6d4ee54ba3041b1458dcaeccb8b11c2c6d324175",
            "sum": -155958.32397764642,
            "top1": 644
          },
          "net_j": {
            "core": 6.023795146479088,
            "dram": 0.7691783647078411,
            "package-0": 7.6593051094554605,
            "psys": 12.52695476994197,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 265546,
          "npu_utilization": 0.38817020252266426,
          "seconds": 0.684096817,
          "watts": {
            "core": 12.577228524073076,
            "dram": 1.6181803108725763,
            "package-0": 16.922949664886396,
            "psys": 28.468083049127824,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 120.8069033498919,
          "decode_tokens_per_s": 3.914922688470836,
          "end_to_end_ms": 67810.694363,
          "handoff_estimate_ms": 459.60501500000004,
          "handoff_plus_first_decode_ms": 684.096817,
          "output_tokens_per_s_e2e": 0.2359509831052576,
          "prefill_tokens_per_s": 127.791530098787,
          "steady_tpot_ms": 224.49180199999998,
          "steady_tpot_p95_ms": 226.46343520000002,
          "ttft_ms": 63979.201076000005
        },
        "prefill": {
          "gpu_active_fraction": 0.99857120177806,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6299,
          "gross_j": {
            "core": 364.959943,
            "dram": 147.338184,
            "package-0": 1262.941347,
            "psys": 2541.350367,
            "uncore": 517.074421
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 123.64595327691865,
            "dram": 115.74463472235213,
            "package-0": 896.5503797289184,
            "psys": 1891.551137750049,
            "uncore": 516.9494647944705
          },
          "npu_busy_delta_us": 19467,
          "npu_utilization": 0.00030427075788075917,
          "seconds": 63.979201076,
          "watts": {
            "core": 5.704352928172223,
            "dram": 2.302907531230017,
            "package-0": 19.739873673942405,
            "psys": 39.72150830675684,
            "uncore": 8.081914314399995
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1793.833741861425,
            "sha256": "abce275f1b23bba1147d8eaa404898b4767f46363c829f7f936abbf8b0cc514d",
            "sum": -631544.1516896337,
            "top1": 15
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 309,
          "gross_j": {
            "core": 15.368613,
            "dram": 9.694494,
            "package-0": 48.098875,
            "psys": 87.459554,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.4973988140987053,
            "dram": 8.14027913331044,
            "package-0": 30.074616840554654,
            "psys": 55.493293622341014,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3129592,
          "npu_utilization": 0.9943431117847064,
          "seconds": 3.14739647,
          "token_latencies_s": [
            0.228600032,
            0.22531296,
            0.225197055,
            0.224492072,
            0.224491532,
            0.224765094,
            0.224612395,
            0.224616439,
            0.224083261,
            0.224131079,
            0.224280993,
            0.224314108,
            0.224252064,
            0.224197104
          ],
          "watts": {
            "core": 4.882960614110367,
            "dram": 3.0801629513170297,
            "package-0": 15.282115061913379,
            "psys": 27.787904966418164,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 405.725036,
            "dram": 157.771631,
            "package-0": 1335.555419,
            "psys": 2675.2137940000002,
            "uncore": 513.700101
          },
          "net_j": {
            "core": 151.32154861726053,
            "dram": 124.46436553952897,
            "package-0": 949.290453896807,
            "psys": 1990.16776658358,
            "uncore": 513.5753062151273
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 40,
          "gross_j": {
            "core": 4.874621,
            "dram": 0.902342,
            "package-0": 7.28941,
            "psys": 12.405242,
            "uncore": 0.0
          },
          "logits": {
            "l2": 989.2704500041049,
            "sha256": "cb3068c96c42c3db93aa7cf355e8b3d774627ba1dca0013bcfc7356782d3aa5c",
            "sum": -167179.82651147246,
            "top1": 644
          },
          "net_j": {
            "core": 3.3163420704470203,
            "dram": 0.6983274619052634,
            "package-0": 4.9234496802844285,
            "psys": 8.209180062016493,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 253173,
          "npu_utilization": 0.6127959366657525,
          "seconds": 0.413144058,
          "watts": {
            "core": 11.798840877919634,
            "dram": 2.184085629521507,
            "package-0": 17.64374885430399,
            "psys": 30.02643208776344,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 121.45364990229947,
          "decode_tokens_per_s": 4.221727922177888,
          "end_to_end_ms": 67449.59914,
          "handoff_estimate_ms": 189.06075549999997,
          "handoff_plus_first_decode_ms": 413.144058,
          "output_tokens_per_s_e2e": 0.23721415996542866,
          "prefill_tokens_per_s": 127.95682701199303,
          "steady_tpot_ms": 224.0833025,
          "steady_tpot_p95_ms": 225.22429145,
          "ttft_ms": 63896.551602
        },
        "prefill": {
          "gpu_active_fraction": 0.9982520260607024,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6293,
          "gross_j": {
            "core": 384.148187,
            "dram": 147.189443,
            "package-0": 1278.945456,
            "psys": 2573.526615,
            "uncore": 513.700101
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 143.1459309963779,
            "dram": 115.63670683326436,
            "package-0": 913.0277991206399,
            "psys": 1924.5668079935267,
            "uncore": 513.5753062151273
          },
          "npu_busy_delta_us": 18955,
          "npu_utilization": 0.00029665137671383655,
          "seconds": 63.896551602,
          "watts": {
            "core": 6.012033159360292,
            "dram": 2.303558475531141,
            "package-0": 20.01587603610158,
            "psys": 40.276455465547336,
            "uncore": 8.039559070413448
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1742.1929938026972,
            "sha256": "e2843afb7cd2280dfd2dae620c6d9141e0428f52bd1f1c78f70eb6d15dc51199",
            "sum": -606370.0193425957,
            "top1": 15
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 309,
          "gross_j": {
            "core": 16.702228,
            "dram": 9.679846,
            "package-0": 49.320553,
            "psys": 89.281937,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.859275550435603,
            "dram": 8.129331244359351,
            "package-0": 31.339205095882647,
            "psys": 57.391778528036795,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3125576,
          "npu_utilization": 0.9954369680178832,
          "seconds": 3.13990348,
          "token_latencies_s": [
            0.225399757,
            0.22439506,
            0.224244435,
            0.224206925,
            0.224036196,
            0.224085891,
            0.223933938,
            0.224080714,
            0.223920374,
            0.223989555,
            0.223982291,
            0.224072247,
            0.22512981,
            0.224385639
          ],
          "watts": {
            "core": 5.31934440226806,
            "dram": 3.082848266405947,
            "package-0": 15.70766531969957,
            "psys": 28.434611945460183,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 384.387932,
            "dram": 151.061503,
            "package-0": 1251.380952,
            "psys": 2532.750842,
            "uncore": 465.706999
          },
          "net_j": {
            "core": 135.3611980303917,
            "dram": 118.45817817764305,
            "package-0": 873.2795993176032,
            "psys": 1862.1830892485877,
            "uncore": 465.58499751665295
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 41,
          "gross_j": {
            "core": 5.410326,
            "dram": 0.902525,
            "package-0": 7.822673,
            "psys": 13.118253,
            "uncore": 0.0
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 3.8428971579684044,
            "dram": 0.6973125279972001,
            "package-0": 5.442820218901095,
            "psys": 8.897552597901203,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 253201,
          "npu_utilization": 0.6092860995053637,
          "seconds": 0.41556996,
          "watts": {
            "core": 13.019049788873094,
            "dram": 2.171776323774702,
            "package-0": 18.823961674226886,
            "psys": 31.566894296209473,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 124.07596404600552,
          "decode_tokens_per_s": 4.216178673446935,
          "end_to_end_ms": 66024.068908,
          "handoff_estimate_ms": 191.29923050000002,
          "handoff_plus_first_decode_ms": 415.56996000000004,
          "output_tokens_per_s_e2e": 0.24233586727735454,
          "prefill_tokens_per_s": 130.8864799171463,
          "steady_tpot_ms": 224.2707295,
          "steady_tpot_p95_ms": 225.40313005000002,
          "ttft_ms": 62466.344921
        },
        "prefill": {
          "gpu_active_fraction": 0.9980456026058632,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 6140,
          "gross_j": {
            "core": 364.317732,
            "dram": 140.47132,
            "package-0": 1196.604908,
            "psys": 2433.9248,
            "uncore": 465.706999
          },
          "logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "net_j": {
            "core": 128.7098678380444,
            "dram": 109.62483374719685,
            "package-0": 838.8776442552257,
            "psys": 1799.4907636088546,
            "uncore": 465.58499751665295
          },
          "npu_busy_delta_us": 17037,
          "npu_utilization": 0.00027273886476864256,
          "seconds": 62.466344921,
          "watts": {
            "core": 5.8322242554890265,
            "dram": 2.2487520308359867,
            "package-0": 19.155993671685504,
            "psys": 38.96377806446236,
            "uncore": 7.45532653765753
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": null,
            "sha256": "7b769dbc741082009e7e9a93ca58b6e7f8d647e22fffd35e2edcda73c4a3253b",
            "sum": null,
            "top1": 0
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 309,
          "gross_j": {
            "core": 14.659874,
            "dram": 9.687658,
            "package-0": 46.953371,
            "psys": 85.707789,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.808433034378888,
            "dram": 8.136031902449007,
            "package-0": 28.959134843476377,
            "psys": 53.794773041831974,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3125760,
          "npu_utilization": 0.9947825514411042,
          "seconds": 3.142154027,
          "token_latencies_s": [
            0.226267502,
            0.224937699,
            0.224006437,
            0.224226763,
            0.224302905,
            0.22430406,
            0.224367411,
            0.224536623,
            0.224238554,
            0.223990205,
            0.224109749,
            0.224024072,
            0.22414712,
            0.224649712
          ],
          "watts": {
            "core": 4.665549134138611,
            "dram": 3.0831263893353373,
            "package-0": 14.94305199443999,
            "psys": 27.276762457704944,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 6.129528114572167,
    "compile_target": "AUTO:NPU,GPU",
    "context_length": 8192,
    "devices": {
      "GPU": {
        "DEVICE_ARCHITECTURE": "GPU: vendor=0x8086 arch=v30.1.0",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) Graphics (iGPU)"
      },
      "NPU": {
        "DEVICE_ARCHITECTURE": "5010",
        "DEVICE_TYPE": "Type.INTEGRATED",
        "FULL_DEVICE_NAME": "Intel(R) AI Boost"
      }
    },
    "energy_domains": [
      "package-0",
      "core",
      "uncore",
      "dram",
      "psys"
    ],
    "execution_devices": [
      "GPU.0"
    ],
    "host": {
      "hostname": "wuklab-NUC16",
      "pid": 401237,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 3.7717568469857548,
      "dram": 0.4938096873094483,
      "package-0": 5.7267199513143465,
      "psys": 10.156413620702507,
      "uncore": 0.0019530754280777179
    },
    "model": {
      "family": "Qwen2.5",
      "ir": "/data/models/qwen25_7b_ov_fp16/openvino_model.xml",
      "key": "qwen25_7b_fp16",
      "size_b": 7.0
    },
    "ok": true,
    "openvino_version": "2026.3.0-22451-8a17657b995-releases/2026/3",
    "output_tokens": 16,
    "placement": {
      "decode_device": null,
      "key": "openvino_auto",
      "label": "OpenVINO AUTO",
      "mode": "stock",
      "prefill_device": null
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "PERFORMANCE_HINT": "LATENCY"
    },
    "read_model_s": 0.04655962623655796,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 95.417542,
            "dram": 41.425066,
            "package-0": 470.951858,
            "psys": 788.19183,
            "uncore": 286.679137
          },
          "net_j": {
            "core": 33.85747181940967,
            "dram": 33.365437492685516,
            "package-0": 377.48419918492226,
            "psys": 622.4257001464254,
            "uncore": 286.6472602207436
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9166666666666666,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 24,
          "gross_j": {
            "core": 3.162224,
            "dram": 0.687926,
            "package-0": 6.386031,
            "psys": 10.566563,
            "uncore": 1.796749
          },
          "logits": {
            "l2": 842.3149968162743,
            "sha256": "d805104637e3082ed7ef6942f1d19e2c0a800de34252afd57ab996e71250bf5f",
            "sum": -27488.31816223578,
            "top1": 78
          },
          "net_j": {
            "core": 2.265249997072613,
            "dram": 0.570491480788017,
            "package-0": 5.024140738574577,
            "psys": 8.151232598193172,
            "uncore": 1.7962845326560506
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.237813316,
          "watts": {
            "core": 13.297085517280287,
            "dram": 2.8927143844207617,
            "package-0": 26.853126256395164,
            "psys": 44.43217553049048,
            "uncore": 7.555291815534837
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 501.9200270543129,
          "decode_tokens_per_s": 5.0185961621677,
          "end_to_end_ms": 16321.325228,
          "handoff_estimate_ms": 41.48733100000002,
          "handoff_plus_first_decode_ms": 237.813316,
          "output_tokens_per_s_e2e": 0.9803125528404549,
          "prefill_tokens_per_s": 613.2410144329137,
          "steady_tpot_ms": 196.32598499999997,
          "steady_tpot_p95_ms": 197.32361790000002,
          "ttft_ms": 13332.441581
        },
        "prefill": {
          "gpu_active_fraction": 0.9604863221884499,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1316,
          "gross_j": {
            "core": 72.611569,
            "dram": 31.694499,
            "package-0": 400.913158,
            "psys": 663.302989,
            "uncore": 259.087532
          },
          "logits": {
            "l2": 1025.1392851150267,
            "sha256": "4ff86f4268bae4a3871f14f73062d317d9c93b8e3d5513020b94dee3023132a8",
            "sum": -317470.8328210273,
            "top1": 8
          },
          "net_j": {
            "core": 22.324841179825675,
            "dram": 25.110810191814906,
            "package-0": 324.56199879835435,
            "psys": 527.8931977295111,
            "uncore": 259.0614927359519
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 13.332441581,
          "watts": {
            "core": 5.446231926752142,
            "dram": 2.377246418628054,
            "package-0": 30.07049800775722,
            "psys": 49.751051596226006,
            "uncore": 19.432864597676126
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 727.8313868092002,
            "sha256": "474801b73ad22489ab599f54a4d0eb647b9530fa6cbcb6caa0057490e9977584",
            "sum": -118284.7835280532,
            "top1": 42
          },
          "gpu_active_fraction": 0.9963235294117647,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 272,
          "gross_j": {
            "core": 19.643749,
            "dram": 9.042641,
            "package-0": 63.652669,
            "psys": 114.322278,
            "uncore": 25.794856
          },
          "net_j": {
            "core": 9.267380642511382,
            "dram": 7.684135820082589,
            "package-0": 47.898059647993335,
            "psys": 86.38126981872104,
            "uncore": 25.78948295213561
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.751070331,
          "token_latencies_s": [
            0.198835878,
            0.19628435,
            0.196502363,
            0.196204352,
            0.196407612,
            0.196509324,
            0.196200959,
            0.19615562,
            0.196468163,
            0.196266558,
            0.196319751,
            0.196047689,
            0.196503537,
            0.196332219
          ],
          "watts": {
            "core": 7.140402329467017,
            "dram": 3.2869537714483092,
            "package-0": 23.137419746322,
            "psys": 41.55556356076307,
            "uncore": 9.376298275378405
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 95.01428399999999,
            "dram": 41.655656,
            "package-0": 471.34614400000004,
            "psys": 780.6806190000001,
            "uncore": 287.188657
          },
          "net_j": {
            "core": 32.700097034429604,
            "dram": 33.49729626771712,
            "package-0": 376.73349736800674,
            "psys": 612.88383810043,
            "uncore": 287.1563897271294
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9545454545454546,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 2.889397,
            "dram": 0.679625,
            "package-0": 6.096908,
            "psys": 10.03037,
            "uncore": 1.799922
          },
          "logits": {
            "l2": 842.3149968162743,
            "sha256": "d805104637e3082ed7ef6942f1d19e2c0a800de34252afd57ab996e71250bf5f",
            "sum": -27488.31816223578,
            "top1": 78
          },
          "net_j": {
            "core": 2.009652838732223,
            "dram": 0.5644462631906187,
            "package-0": 4.761178088245295,
            "psys": 7.661435324538587,
            "uncore": 1.7994664545409282
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.233245195,
          "watts": {
            "core": 12.387809318001173,
            "dram": 2.913779209899694,
            "package-0": 26.139479529256757,
            "psys": 43.00354397439999,
            "uncore": 7.716866364599708
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 495.84586745196697,
          "decode_tokens_per_s": 5.011159632009462,
          "end_to_end_ms": 16521.263033,
          "handoff_estimate_ms": 36.44838249999999,
          "handoff_plus_first_decode_ms": 233.245195,
          "output_tokens_per_s_e2e": 0.968448959867123,
          "prefill_tokens_per_s": 604.3786150972744,
          "steady_tpot_ms": 196.7968125,
          "steady_tpot_p95_ms": 199.01717010000002,
          "ttft_ms": 13527.943901
        },
        "prefill": {
          "gpu_active_fraction": 0.9506357516828721,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1337,
          "gross_j": {
            "core": 70.247745,
            "dram": 31.93516,
            "package-0": 399.082962,
            "psys": 655.58725,
            "uncore": 259.349922
          },
          "logits": {
            "l2": 1025.1392851150267,
            "sha256": "4ff86f4268bae4a3871f14f73062d317d9c93b8e3d5513020b94dee3023132a8",
            "sum": -317470.8328210273,
            "top1": 8
          },
          "net_j": {
            "core": 19.22362996576406,
            "dram": 25.25493025230743,
            "package-0": 321.61221576188206,
            "psys": 518.1918563037842,
            "uncore": 259.3235009051745
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 13.527943901,
          "watts": {
            "core": 5.192788018200401,
            "dram": 2.360680989935161,
            "package-0": 29.50063697192737,
            "psys": 48.46170673072782,
            "uncore": 19.171422050384802
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 727.8313868092002,
            "sha256": "474801b73ad22489ab599f54a4d0eb647b9530fa6cbcb6caa0057490e9977584",
            "sum": -118284.7835280532,
            "top1": 42
          },
          "gpu_active_fraction": 0.9963369963369964,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 273,
          "gross_j": {
            "core": 21.877142,
            "dram": 9.040871,
            "package-0": 66.166274,
            "psys": 115.062999,
            "uncore": 26.038813
          },
          "net_j": {
            "core": 11.46681422993332,
            "dram": 7.677919752219071,
            "package-0": 50.36010351787937,
            "psys": 87.0305464721072,
            "uncore": 26.03342236741397
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.760073937,
          "token_latencies_s": [
            0.201836562,
            0.197499036,
            0.197044282,
            0.196866754,
            0.196715114,
            0.196646921,
            0.19685332,
            0.196865295,
            0.196312369,
            0.196651672,
            0.196740305,
            0.196678852,
            0.196972966,
            0.196358737
          ],
          "watts": {
            "core": 7.926288389136004,
            "dram": 3.2755901495257658,
            "package-0": 23.972645483518438,
            "psys": 41.68837561107697,
            "uncore": 9.434099808319736
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 96.90227900000001,
            "dram": 41.682266,
            "package-0": 474.959294,
            "psys": 790.577383,
            "uncore": 288.63939799999997
          },
          "net_j": {
            "core": 34.54050157181344,
            "dram": 33.517675582013666,
            "package-0": 380.27438999224984,
            "psys": 622.6524526882756,
            "uncore": 288.60710608403417
          }
        },
        "first_decode": {
          "gpu_active_fraction": 0.9090909090909091,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 22,
          "gross_j": {
            "core": 3.508414,
            "dram": 0.681578,
            "package-0": 6.713057,
            "psys": 10.765841,
            "uncore": 1.788814
          },
          "logits": {
            "l2": 842.3149968162743,
            "sha256": "d805104637e3082ed7ef6942f1d19e2c0a800de34252afd57ab996e71250bf5f",
            "sum": -27488.31816223578,
            "top1": 78
          },
          "net_j": {
            "core": 2.6371966645232305,
            "dram": 0.567515620725958,
            "package-0": 5.3902735069461105,
            "psys": 8.419866969436566,
            "uncore": 1.7883628698663607
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 0.230984491,
          "watts": {
            "core": 15.18895915830124,
            "dram": 2.95075222171518,
            "package-0": 29.06280404774016,
            "psys": 46.60850152056313,
            "uncore": 7.744303490921388
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 495.4674700554924,
          "decode_tokens_per_s": 5.0314153237436505,
          "end_to_end_ms": 16533.880618,
          "handoff_estimate_ms": 34.78480849999998,
          "handoff_plus_first_decode_ms": 230.984491,
          "output_tokens_per_s_e2e": 0.9677099024521336,
          "prefill_tokens_per_s": 603.2785360492352,
          "steady_tpot_ms": 196.1996825,
          "steady_tpot_p95_ms": 197.8425613,
          "ttft_ms": 13552.612121
        },
        "prefill": {
          "gpu_active_fraction": 0.9507462686567164,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 1340,
          "gross_j": {
            "core": 76.778429,
            "dram": 31.930155,
            "package-0": 407.529169,
            "psys": 670.526787,
            "uncore": 260.990238
          },
          "logits": {
            "l2": 1025.1392851150267,
            "sha256": "4ff86f4268bae4a3871f14f73062d317d9c93b8e3d5513020b94dee3023132a8",
            "sum": -317470.8328210273,
            "top1": 8
          },
          "net_j": {
            "core": 25.66127143807612,
            "dram": 25.23774384630275,
            "package-0": 329.9171547742447,
            "psys": 532.8808526581777,
            "uncore": 260.9637687262802
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 13.552612121,
          "watts": {
            "core": 5.665212603630155,
            "dram": 2.356014819499164,
            "package-0": 30.070156613463965,
            "psys": 49.47583395831181,
            "uncore": 19.2575597729674
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 727.8313868092002,
            "sha256": "474801b73ad22489ab599f54a4d0eb647b9530fa6cbcb6caa0057490e9977584",
            "sum": -118284.7835280532,
            "top1": 42
          },
          "gpu_active_fraction": 0.9963235294117647,
          "gpu_freq_mhz_after": 2450,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 2450,
          "gpu_sample_count": 272,
          "gross_j": {
            "core": 16.615436,
            "dram": 9.070533,
            "package-0": 60.717068,
            "psys": 109.284755,
            "uncore": 25.860346
          },
          "net_j": {
            "core": 6.242033469214089,
            "dram": 7.712416114984963,
            "package-0": 44.96696171105906,
            "psys": 81.35173306066136,
            "uncore": 25.854974487887645
          },
          "npu_busy_delta_us": 0,
          "npu_utilization": 0.0,
          "seconds": 2.750284006,
          "token_latencies_s": [
            0.200131574,
            0.196610016,
            0.19598051,
            0.196225125,
            0.1963334,
            0.196521574,
            0.196343023,
            0.195932347,
            0.196058054,
            0.196261632,
            0.195909876,
            0.19617424,
            0.196118327,
            0.195653409
          ],
          "watts": {
            "core": 6.041352807110787,
            "dram": 3.298035032095518,
            "package-0": 22.07665385376204,
            "psys": 39.73580719721497,
            "uncore": 9.402791109421155
          }
        }
      }
    ]
  }
]
