[
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 6.715313421562314,
    "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": 408986,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "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": 0.025821217335760593,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 7.257427,
            "dram": 2.338556,
            "package-0": 18.084488,
            "psys": 30.446029000000003,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.718181396657727,
            "dram": 1.8473279441429187,
            "package-0": 14.807754652818453,
            "psys": 22.366153392982568,
            "uncore": 0.0
          }
        },
        "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": 2.491815,
            "dram": 0.163513,
            "package-0": 3.020195,
            "psys": 4.277272,
            "uncore": 0.0
          },
          "logits": {
            "l2": 403.47644126655484,
            "sha256": "82fc2ee3733521e21ac618cc56a1a447e6bfe4441ae27196716ae874e9aa61fb",
            "sum": 41930.551351014525,
            "top1": 546
          },
          "net_j": {
            "core": 2.330040537078452,
            "dram": 0.11188501111003939,
            "package-0": 2.675810857387284,
            "psys": 3.4280783859595076,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 39835,
          "npu_utilization": 0.3757458439453512,
          "seconds": 0.10601581,
          "watts": {
            "core": 23.504183008175854,
            "dram": 1.5423454294222718,
            "package-0": 28.48815662494113,
            "psys": 40.34560505645337,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1015.1526804501798,
          "decode_tokens_per_s": 23.534573459649497,
          "end_to_end_ms": 1008.7152600000001,
          "handoff_estimate_ms": 68.2995795,
          "handoff_plus_first_decode_ms": 106.01581,
          "output_tokens_per_s_e2e": 15.86176063203406,
          "prefill_tokens_per_s": 2714.383233196603,
          "steady_tpot_ms": 37.7162305,
          "steady_tpot_p95_ms": 38.885700449999995,
          "ttft_ms": 371.355079
        },
        "prefill": {
          "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": 0.991452,
            "dram": 0.557799,
            "package-0": 6.010361,
            "psys": 10.145238,
            "uncore": 0.0
          },
          "logits": {
            "l2": 628.0526414006035,
            "sha256": "701f4560c1677db1b1d9f45ca56d23a3429b2db7c259cf40c52b2ab0d3b85d4b",
            "sum": -187921.24455487914,
            "top1": 5809
          },
          "net_j": {
            "core": 0.42478398644226645,
            "dram": 0.37695506839354964,
            "package-0": 4.804042805701504,
            "psys": 7.17065938169484,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 366875,
          "npu_utilization": 0.987935861784726,
          "seconds": 0.371355079,
          "watts": {
            "core": 2.669822108451626,
            "dram": 1.5020637431486432,
            "package-0": 16.184943575256714,
            "psys": 27.31950786110024,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 515.7828349073951,
            "sha256": "a5e5c388249e29831855bd73b45711046f70e06f6e5b7ebe55336c2dfcae2631",
            "sum": 92985.67104252055,
            "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": 53,
          "gross_j": {
            "core": 3.77416,
            "dram": 1.617244,
            "package-0": 9.053932,
            "psys": 16.023519,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.963356873137008,
            "dram": 1.3584878646393297,
            "package-0": 7.327900989729665,
            "psys": 11.767415625328221,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 520848,
          "npu_utilization": 0.9802456343326915,
          "seconds": 0.531344371,
          "token_latencies_s": [
            0.038971297,
            0.03883961,
            0.037966017,
            0.038545887,
            0.037585356,
            0.037704661,
            0.0376612,
            0.037537949,
            0.03772751,
            0.037843312,
            0.038023015,
            0.037704951,
            0.037586166,
            0.037621356
          ],
          "watts": {
            "core": 7.103039395894909,
            "dram": 3.0436833215270855,
            "package-0": 17.039668610698428,
            "psys": 30.15656112032097,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 5.938949,
            "dram": 2.293145,
            "package-0": 16.629473,
            "psys": 28.042774,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.491686849301956,
            "dram": 1.831272137697043,
            "package-0": 13.548553268276429,
            "psys": 20.44574197145735,
            "uncore": 0.0
          }
        },
        "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": 4,
          "gross_j": {
            "core": 1.322628,
            "dram": 0.118713,
            "package-0": 1.708247,
            "psys": 2.344537,
            "uncore": 0.0
          },
          "logits": {
            "l2": 403.47644126655484,
            "sha256": "82fc2ee3733521e21ac618cc56a1a447e6bfe4441ae27196716ae874e9aa61fb",
            "sum": 41930.551351014525,
            "top1": 546
          },
          "net_j": {
            "core": 1.2589365598608067,
            "dram": 0.09838679458718103,
            "package-0": 1.5726611874558898,
            "psys": 2.010205095826261,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 37245,
          "npu_utilization": 0.8923315335205556,
          "seconds": 0.041738971,
          "watts": {
            "core": 31.688083541877443,
            "dram": 2.8441764891616517,
            "package-0": 40.9269073739264,
            "psys": 56.1714135214306,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1079.6726075855745,
          "decode_tokens_per_s": 26.146813717555577,
          "end_to_end_ms": 948.435658,
          "handoff_estimate_ms": 3.849347000000003,
          "handoff_plus_first_decode_ms": 41.738971,
          "output_tokens_per_s_e2e": 16.8698844935246,
          "prefill_tokens_per_s": 2689.7789121813603,
          "steady_tpot_ms": 37.889624,
          "steady_tpot_p95_ms": 38.696261199999995,
          "ttft_ms": 374.75199
        },
        "prefill": {
          "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": 0.90344,
            "dram": 0.558531,
            "package-0": 5.928146,
            "psys": 9.847509,
            "uncore": 0.0
          },
          "logits": {
            "l2": 628.0526414006035,
            "sha256": "701f4560c1677db1b1d9f45ca56d23a3429b2db7c259cf40c52b2ab0d3b85d4b",
            "sum": -187921.24455487914,
            "top1": 5809
          },
          "net_j": {
            "core": 0.33158848171200694,
            "dram": 0.37603282754618744,
            "package-0": 4.710793203576909,
            "psys": 6.845720901064885,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 371268,
          "npu_utilization": 0.9907032114759418,
          "seconds": 0.37475199,
          "watts": {
            "core": 2.4107677186717544,
            "dram": 1.490401692063063,
            "package-0": 15.818851288821708,
            "psys": 26.27740282313111,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 515.7828349073951,
            "sha256": "a5e5c388249e29831855bd73b45711046f70e06f6e5b7ebe55336c2dfcae2631",
            "sum": 92985.67104252055,
            "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": 53,
          "gross_j": {
            "core": 3.712881,
            "dram": 1.615901,
            "package-0": 8.99308,
            "psys": 15.850728,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.9011618077291423,
            "dram": 1.3568525155636744,
            "package-0": 7.265098877243631,
            "psys": 11.589815974566203,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 519663,
          "npu_utilization": 0.9769117032855766,
          "seconds": 0.531944697,
          "token_latencies_s": [
            0.038830624,
            0.038623912,
            0.037920425,
            0.037767377,
            0.037724127,
            0.037836459,
            0.037921237,
            0.037796923,
            0.037858823,
            0.037717425,
            0.037835252,
            0.037922573,
            0.038166028,
            0.037974961
          ],
          "watts": {
            "core": 6.979825197881425,
            "dram": 3.037723675248895,
            "package-0": 16.906043148316225,
            "psys": 29.797699064194262,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.222091000000001,
            "dram": 2.299615,
            "package-0": 16.932331,
            "psys": 28.363757,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.772175370108659,
            "dram": 1.8368953180455525,
            "package-0": 13.845762563651352,
            "psys": 20.752796211965624,
            "uncore": 0.0
          }
        },
        "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": 5,
          "gross_j": {
            "core": 1.419979,
            "dram": 0.123535,
            "package-0": 1.821956,
            "psys": 2.481256,
            "uncore": 0.0
          },
          "logits": {
            "l2": 403.47644126655484,
            "sha256": "82fc2ee3733521e21ac618cc56a1a447e6bfe4441ae27196716ae874e9aa61fb",
            "sum": 41930.551351014525,
            "top1": 546
          },
          "net_j": {
            "core": 1.351889310293606,
            "dram": 0.10180515661064307,
            "package-0": 1.6770072300468963,
            "psys": 2.1238366110776052,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 37224,
          "npu_utilization": 0.8342208022719205,
          "seconds": 0.04462128,
          "watts": {
            "core": 31.82291050368793,
            "dram": 2.7685221042516037,
            "package-0": 40.83154943112344,
            "psys": 55.60701082532819,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1077.6967072361067,
          "decode_tokens_per_s": 25.956409774464063,
          "end_to_end_ms": 950.1745649999999,
          "handoff_estimate_ms": 6.635176999999999,
          "handoff_plus_first_decode_ms": 44.62128,
          "output_tokens_per_s_e2e": 16.839011050564167,
          "prefill_tokens_per_s": 2707.6203193742926,
          "steady_tpot_ms": 37.986103,
          "steady_tpot_p95_ms": 38.633267800000006,
          "ttft_ms": 372.282625
        },
        "prefill": {
          "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": 1.010984,
            "dram": 0.558897,
            "package-0": 6.023972,
            "psys": 9.9791,
            "uncore": 0.0
          },
          "logits": {
            "l2": 628.0526414006035,
            "sha256": "701f4560c1677db1b1d9f45ca56d23a3429b2db7c259cf40c52b2ab0d3b85d4b",
            "sum": -187921.24455487914,
            "top1": 5809
          },
          "net_j": {
            "core": 0.44290060079379556,
            "dram": 0.37760136845763503,
            "package-0": 4.814640744404855,
            "psys": 6.9970916772014355,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 367567,
          "npu_utilization": 0.987333212233582,
          "seconds": 0.372282625,
          "watts": {
            "core": 2.715635735081647,
            "dram": 1.5012707079735457,
            "package-0": 16.181179554108926,
            "psys": 26.805172548678577,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 515.7828349073951,
            "sha256": "a5e5c388249e29831855bd73b45711046f70e06f6e5b7ebe55336c2dfcae2631",
            "sum": 92985.67104252055,
            "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": 53,
          "gross_j": {
            "core": 3.791128,
            "dram": 1.617183,
            "package-0": 9.086403,
            "psys": 15.903401,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.977385459021258,
            "dram": 1.3574887929772743,
            "package-0": 7.3541145891996,
            "psys": 11.631867923686585,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 520296,
          "npu_utilization": 0.9756696533801428,
          "seconds": 0.53327066,
          "token_latencies_s": [
            0.038574199,
            0.038742967,
            0.038039353,
            0.0379759,
            0.038109088,
            0.037935448,
            0.037874742,
            0.038142298,
            0.037996306,
            0.037838648,
            0.037940378,
            0.037896361,
            0.037962355,
            0.03819538
          ],
          "watts": {
            "core": 7.1092004199143455,
            "dram": 3.032574490409804,
            "package-0": 17.039007921418367,
            "psys": 29.822381377591636,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 14.859274039976299,
    "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": 409111,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.02583260927349329,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 8.978614,
            "dram": 2.921929,
            "package-0": 26.566399,
            "psys": 43.076733000000004,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.875950641690203,
            "dram": 2.2508942857336987,
            "package-0": 22.09026658220391,
            "psys": 32.03934017478309,
            "uncore": 0.0
          }
        },
        "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": 2.564385,
            "dram": 0.172119,
            "package-0": 3.118522,
            "psys": 4.519947,
            "uncore": 0.0
          },
          "logits": {
            "l2": 392.44545785960474,
            "sha256": "da8a47e01f0c4f258fd33ddaf9c7ae00cfdff7bbd7a3dcc5f722a75b5d939ae9",
            "sum": 47131.24965239875,
            "top1": 3862
          },
          "net_j": {
            "core": 2.393072971142117,
            "dram": 0.11744723436489787,
            "package-0": 2.753834365672465,
            "psys": 3.620688376066361,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 40347,
          "npu_utilization": 0.35938729556691,
          "seconds": 0.112266072,
          "watts": {
            "core": 22.842030137119256,
            "dram": 1.533134605439834,
            "package-0": 27.777955925989822,
            "psys": 40.261023829176104,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1486.2762447718853,
          "decode_tokens_per_s": 22.261778896810455,
          "end_to_end_ms": 1377.940344,
          "handoff_estimate_ms": 72.4623745,
          "handoff_plus_first_decode_ms": 112.266072,
          "output_tokens_per_s_e2e": 11.611533162280354,
          "prefill_tokens_per_s": 2885.79081098679,
          "steady_tpot_ms": 39.8036975,
          "steady_tpot_p95_ms": 41.9399647,
          "ttft_ms": 704.1397430000001
        },
        "prefill": {
          "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": 2.140863,
            "dram": 1.07446,
            "package-0": 13.678982,
            "psys": 21.48194,
            "uncore": 0.0
          },
          "logits": {
            "l2": 443.1691514737982,
            "sha256": "4995d5054aceead0555f3f6020af75f0389032b26365d691afaab466921c7c37",
            "sum": -56908.99897800945,
            "top1": 8
          },
          "net_j": {
            "core": 1.0663833658252293,
            "dram": 0.7315553065530158,
            "package-0": 11.391638615342623,
            "psys": 15.841734325509437,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 685860,
          "npu_utilization": 0.9740396090666338,
          "seconds": 0.704139743,
          "watts": {
            "core": 3.0403950654437066,
            "dram": 1.5259186982149933,
            "package-0": 19.426516023254774,
            "psys": 30.50806351091022,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 453.91327201082566,
            "sha256": "e94495b9cabc339c33632c206c7636bb049ae1092943ac3356384433083bc8a8",
            "sum": 9498.106287172064,
            "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": 56,
          "gross_j": {
            "core": 4.273366,
            "dram": 1.67535,
            "package-0": 9.768895,
            "psys": 17.074846,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.416494304722857,
            "dram": 1.4018917448157848,
            "package-0": 7.94479360118882,
            "psys": 12.576917473207292,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 549591,
          "npu_utilization": 0.9787305528276785,
          "seconds": 0.561534529,
          "token_latencies_s": [
            0.042701948,
            0.040468577,
            0.03970427,
            0.039797149,
            0.039703644,
            0.039811891,
            0.039641551,
            0.041529666,
            0.03946634,
            0.039810246,
            0.039702334,
            0.039452351,
            0.039886598,
            0.039829901
          ],
          "watts": {
            "core": 7.610157130693561,
            "dram": 2.9835208940463924,
            "package-0": 17.396784161068037,
            "psys": 30.407472948115004,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 9.011207,
            "dram": 2.929984,
            "package-0": 26.477105,
            "psys": 42.398208000000004,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.970217923886795,
            "dram": 2.278631744253657,
            "package-0": 22.132264285554072,
            "psys": 31.684558524914166,
            "uncore": 0.0
          }
        },
        "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": 8,
          "gross_j": {
            "core": 2.282526,
            "dram": 0.149047,
            "package-0": 2.761773,
            "psys": 3.796805,
            "uncore": 0.0
          },
          "logits": {
            "l2": 392.44545785960474,
            "sha256": "da8a47e01f0c4f258fd33ddaf9c7ae00cfdff7bbd7a3dcc5f722a75b5d939ae9",
            "sum": 47131.24965239875,
            "top1": 3862
          },
          "net_j": {
            "core": 2.1696686769564257,
            "dram": 0.11303021147219683,
            "package-0": 2.521523261369895,
            "psys": 3.204389384484811,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 39529,
          "npu_utilization": 0.5344725607379613,
          "seconds": 0.073958895,
          "watts": {
            "core": 30.862089002276196,
            "dram": 2.015268075597939,
            "package-0": 37.34199922808473,
            "psys": 51.33669182050381,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1531.1883031532643,
          "decode_tokens_per_s": 23.642482614634385,
          "end_to_end_ms": 1337.523279,
          "handoff_estimate_ms": 34.0508135,
          "handoff_plus_first_decode_ms": 73.958895,
          "output_tokens_per_s_e2e": 11.962408618384877,
          "prefill_tokens_per_s": 2890.172830784942,
          "steady_tpot_ms": 39.908081499999994,
          "steady_tpot_p95_ms": 40.625068549999995,
          "ttft_ms": 703.072141
        },
        "prefill": {
          "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": 2.691277,
            "dram": 1.107602,
            "package-0": 14.195825,
            "psys": 21.975652,
            "uncore": 0.0
          },
          "logits": {
            "l2": 443.1691514737982,
            "sha256": "4995d5054aceead0555f3f6020af75f0389032b26365d691afaab466921c7c37",
            "sum": -56908.99897800945,
            "top1": 8
          },
          "net_j": {
            "core": 1.6184264694482076,
            "dram": 0.7652172115011071,
            "package-0": 11.911949637701003,
            "psys": 16.343997887778656,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 685677,
          "npu_utilization": 0.9752583839045899,
          "seconds": 0.703072141,
          "watts": {
            "core": 3.8278817251557116,
            "dram": 1.5753746100999326,
            "package-0": 20.191135691721282,
            "psys": 31.256610408063374,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 453.91327201082566,
            "sha256": "e94495b9cabc339c33632c206c7636bb049ae1092943ac3356384433083bc8a8",
            "sum": 9498.106287172064,
            "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": 56,
          "gross_j": {
            "core": 4.037404,
            "dram": 1.673335,
            "package-0": 9.519507,
            "psys": 16.625751,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.182122777482162,
            "dram": 1.4003843212803533,
            "package-0": 7.698791386483176,
            "psys": 12.136171252650698,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 547796,
          "npu_utilization": 0.9773480486865541,
          "seconds": 0.560492243,
          "token_latencies_s": [
            0.040758457,
            0.040553244,
            0.040447942,
            0.039903241,
            0.039844073,
            0.039704511,
            0.039912922,
            0.039817207,
            0.039810117,
            0.039626854,
            0.039889025,
            0.040001374,
            0.039966706,
            0.040220469
          ],
          "watts": {
            "core": 7.203318244673728,
            "dram": 2.985473966675396,
            "package-0": 16.984190448466208,
            "psys": 29.662767340029006,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 8.037578,
            "dram": 2.924308,
            "package-0": 25.467098,
            "psys": 41.432634,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.000539839161551,
            "dram": 2.2742166219362443,
            "package-0": 21.1306679614102,
            "psys": 30.73972384285314,
            "uncore": 0.0
          }
        },
        "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": 8,
          "gross_j": {
            "core": 2.283869,
            "dram": 0.148559,
            "package-0": 2.762932,
            "psys": 3.767447,
            "uncore": 0.0
          },
          "logits": {
            "l2": 392.44545785960474,
            "sha256": "da8a47e01f0c4f258fd33ddaf9c7ae00cfdff7bbd7a3dcc5f722a75b5d939ae9",
            "sum": 47131.24965239875,
            "top1": 3862
          },
          "net_j": {
            "core": 2.17011036745201,
            "dram": 0.11225457153051463,
            "package-0": 2.52076356112944,
            "psys": 3.1703001910590887,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 39461,
          "npu_utilization": 0.5293257903055647,
          "seconds": 0.074549551,
          "watts": {
            "core": 30.635583573132454,
            "dram": 1.9927551273917128,
            "package-0": 37.06168532121676,
            "psys": 50.536146086245374,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1534.1581028220958,
          "decode_tokens_per_s": 23.685173976863762,
          "end_to_end_ms": 1334.934122,
          "handoff_estimate_ms": 34.713420000000006,
          "handoff_plus_first_decode_ms": 74.54955100000001,
          "output_tokens_per_s_e2e": 11.985610178297623,
          "prefill_tokens_per_s": 2896.1275725724354,
          "steady_tpot_ms": 39.836130999999995,
          "steady_tpot_p95_ms": 40.53517215000001,
          "ttft_ms": 701.6265510000001
        },
        "prefill": {
          "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": 1.936458,
            "dram": 1.104977,
            "package-0": 13.428738,
            "psys": 21.299751,
            "uncore": 0.0
          },
          "logits": {
            "l2": 443.1691514737982,
            "sha256": "4995d5054aceead0555f3f6020af75f0389032b26365d691afaab466921c7c37",
            "sum": -56908.99897800945,
            "top1": 8
          },
          "net_j": {
            "core": 0.865813362567873,
            "dram": 0.7632961905063087,
            "package-0": 11.149558524783501,
            "psys": 15.679676158799845,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 682896,
          "npu_utilization": 0.9733041017713708,
          "seconds": 0.701626551,
          "watts": {
            "core": 2.759955416795508,
            "dram": 1.5748791125779389,
            "package-0": 19.139438182407094,
            "psys": 30.357675275603984,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 453.91327201082566,
            "sha256": "e94495b9cabc339c33632c206c7636bb049ae1092943ac3356384433083bc8a8",
            "sum": 9498.106287172064,
            "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": 56,
          "gross_j": {
            "core": 3.817251,
            "dram": 1.670772,
            "package-0": 9.275428,
            "psys": 16.365436,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.964616109141668,
            "dram": 1.3986658598994208,
            "package-0": 7.46034587549726,
            "psys": 11.889747492994207,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 547307,
          "npu_utilization": 0.9795062986299508,
          "seconds": 0.55875802,
          "token_latencies_s": [
            0.040762106,
            0.040412977,
            0.039854143,
            0.039657194,
            0.039522151,
            0.039703482,
            0.040054239,
            0.03978252,
            0.03966908,
            0.039658331,
            0.039858988,
            0.039916003,
            0.039818119,
            0.040055482
          ],
          "watts": {
            "core": 6.831671069347695,
            "dram": 2.9901530540894967,
            "package-0": 16.600080299518567,
            "psys": 29.288950519224763,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 11.439364975318313,
    "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": 409297,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.02579568326473236,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.048434,
            "dram": 4.764087,
            "package-0": 45.430243,
            "psys": 75.806384,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.387809437801369,
            "dram": 3.5958514156287014,
            "package-0": 37.63753561891491,
            "psys": 56.59087281560286,
            "uncore": 0.0
          }
        },
        "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": 15,
          "gross_j": {
            "core": 2.946282,
            "dram": 0.217407,
            "package-0": 3.609183,
            "psys": 5.358567,
            "uncore": 0.0
          },
          "logits": {
            "l2": 388.9912025260317,
            "sha256": "dae6a98394073cc6ade52b2f16f4e8282f7be5042a0c7bf4ba934676eabeec04",
            "sum": 48757.71511605475,
            "top1": 78
          },
          "net_j": {
            "core": 2.7204490747413774,
            "dram": 0.14533567585106066,
            "package-0": 3.1284317310149308,
            "psys": 4.173114892749259,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 43861,
          "npu_utilization": 0.2963675157742507,
          "seconds": 0.147995302,
          "watts": {
            "core": 19.907942753480107,
            "dram": 1.4690128474483601,
            "package-0": 24.38714574872113,
            "psys": 36.20768313307675,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1707.4346451585393,
          "decode_tokens_per_s": 19.95787873726198,
          "end_to_end_ms": 2398.920516,
          "handoff_estimate_ms": 105.0421315,
          "handoff_plus_first_decode_ms": 147.99530199999998,
          "output_tokens_per_s_e2e": 6.669666582650544,
          "prefill_tokens_per_s": 2476.723601352069,
          "steady_tpot_ms": 42.9531705,
          "steady_tpot_p95_ms": 44.112806049999996,
          "ttft_ms": 1647.3376349999999
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 163,
          "gross_j": {
            "core": 4.281483,
            "dram": 2.775505,
            "package-0": 32.174051,
            "psys": 52.893053,
            "uncore": 0.0
          },
          "logits": {
            "l2": 384.67176099469424,
            "sha256": "2a1e5f084e13626a7729e5e9158bdf8240bf1f797678c5c26a29afe86cd5ec20",
            "sum": 17386.329117944464,
            "top1": 72
          },
          "net_j": {
            "core": 1.7677337662528974,
            "dram": 1.9732781517799658,
            "package-0": 26.822802360545804,
            "psys": 39.69777014727674,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1604811,
          "npu_utilization": 0.9741846273062292,
          "seconds": 1.647337635,
          "watts": {
            "core": 2.5990318614920733,
            "dram": 1.6848428282281063,
            "package-0": 19.530939084020865,
            "psys": 32.108204096241636,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 441.38267318488283,
            "sha256": "d3c8964f9838d4f83cf8587ff10448c8cc4639e4ad36480f4b6e161a894e35c1",
            "sum": -5983.314064975828,
            "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": 60,
          "gross_j": {
            "core": 3.820669,
            "dram": 1.771175,
            "package-0": 9.647009,
            "psys": 17.554764,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.899626596807094,
            "dram": 1.477237587997675,
            "package-0": 7.686301527354175,
            "psys": 12.719987775576858,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 589383,
          "npu_utilization": 0.9764664159863369,
          "seconds": 0.603587579,
          "token_latencies_s": [
            0.044939218,
            0.043667815,
            0.043048877,
            0.042954279,
            0.042972128,
            0.043029527,
            0.042938416,
            0.042922471,
            0.042952661,
            0.042695012,
            0.042724293,
            0.042807169,
            0.04294501,
            0.04295368
          ],
          "watts": {
            "core": 6.329933108182797,
            "dram": 2.9344126049353307,
            "package-0": 15.982782508518122,
            "psys": 29.084037860891762,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 11.895233000000001,
            "dram": 4.741383,
            "package-0": 46.125187000000004,
            "psys": 75.202079,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.29823999521884,
            "dram": 3.5934545103337117,
            "package-0": 38.467937953614836,
            "psys": 56.32058538076437,
            "uncore": 0.0
          }
        },
        "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": 2.454095,
            "dram": 0.19397,
            "package-0": 3.033134,
            "psys": 4.391347,
            "uncore": 0.0
          },
          "logits": {
            "l2": 388.9912025260317,
            "sha256": "dae6a98394073cc6ade52b2f16f4e8282f7be5042a0c7bf4ba934676eabeec04",
            "sum": 48757.71511605475,
            "top1": 78
          },
          "net_j": {
            "core": 2.2836638329623327,
            "dram": 0.13957934872279093,
            "package-0": 2.6703215370514135,
            "psys": 3.496712234610351,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 42783,
          "npu_utilization": 0.38305536339466617,
          "seconds": 0.111688816,
          "watts": {
            "core": 21.97261183250434,
            "dram": 1.736700297727214,
            "package-0": 27.157007376638322,
            "psys": 39.31769676920919,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1737.63952059632,
          "decode_tokens_per_s": 21.032105365822613,
          "end_to_end_ms": 2357.220788,
          "handoff_estimate_ms": 68.881819,
          "handoff_plus_first_decode_ms": 111.688816,
          "output_tokens_per_s_e2e": 6.787654377329375,
          "prefill_tokens_per_s": 2481.713442794406,
          "steady_tpot_ms": 42.806997,
          "steady_tpot_p95_ms": 43.764470949999996,
          "ttft_ms": 1644.025426
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 163,
          "gross_j": {
            "core": 5.126818,
            "dram": 2.775506,
            "package-0": 32.946083,
            "psys": 52.935107,
            "uncore": 0.0
          },
          "logits": {
            "l2": 384.67176099469424,
            "sha256": "2a1e5f084e13626a7729e5e9158bdf8240bf1f797678c5c26a29afe86cd5ec20",
            "sum": 17386.329117944464,
            "top1": 72
          },
          "net_j": {
            "core": 2.6181230202557573,
            "dram": 1.9748921441200067,
            "package-0": 27.605593814787767,
            "psys": 39.76635515931322,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1603712,
          "npu_utilization": 0.9754788305810546,
          "seconds": 1.644025426,
          "watts": {
            "core": 3.118454203274591,
            "dram": 1.688237880087385,
            "package-0": 20.039886536402022,
            "psys": 32.19847221511281,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 441.38267318488283,
            "sha256": "d3c8964f9838d4f83cf8587ff10448c8cc4639e4ad36480f4b6e161a894e35c1",
            "sum": -5983.314064975828,
            "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": 60,
          "gross_j": {
            "core": 4.31432,
            "dram": 1.771907,
            "package-0": 10.14597,
            "psys": 17.875625,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.39645314200075,
            "dram": 1.478983017490914,
            "package-0": 8.192022601775658,
            "psys": 13.057517986840796,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 588688,
          "npu_utilization": 0.9786892660017702,
          "seconds": 0.601506546,
          "token_latencies_s": [
            0.044581649,
            0.043324452,
            0.042817876,
            0.042752824,
            0.042770664,
            0.04286276,
            0.042796118,
            0.042584708,
            0.042708335,
            0.042969792,
            0.042929798,
            0.042736589,
            0.042690708,
            0.042946417
          ],
          "watts": {
            "core": 7.172523771669811,
            "dram": 2.945781740503286,
            "package-0": 16.867596982061773,
            "psys": 29.71808888676666,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 12.009611,
            "dram": 4.748646,
            "package-0": 46.410463,
            "psys": 75.039602,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.388613804104695,
            "dram": 3.5930569186075965,
            "package-0": 38.702114029518874,
            "psys": 56.03210452678017,
            "uncore": 0.0
          }
        },
        "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": 2.476006,
            "dram": 0.195679,
            "package-0": 3.065055,
            "psys": 4.396595,
            "uncore": 0.0
          },
          "logits": {
            "l2": 388.9912025260317,
            "sha256": "dae6a98394073cc6ade52b2f16f4e8282f7be5042a0c7bf4ba934676eabeec04",
            "sum": 48757.71511605475,
            "top1": 78
          },
          "net_j": {
            "core": 2.2985648617300547,
            "dram": 0.1390512174082244,
            "package-0": 2.6873197680990697,
            "psys": 3.4651631858206136,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 43042,
          "npu_utilization": 0.37014974897333575,
          "seconds": 0.116282667,
          "watts": {
            "core": 21.292992875713797,
            "dram": 1.6827873409542626,
            "package-0": 26.35865756329789,
            "psys": 37.809547316282305,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1726.1204199499105,
          "decode_tokens_per_s": 20.805680993969492,
          "end_to_end_ms": 2372.951477,
          "handoff_estimate_ms": 73.232411,
          "handoff_plus_first_decode_ms": 116.282667,
          "output_tokens_per_s_e2e": 6.742657890429338,
          "prefill_tokens_per_s": 2469.741815890334,
          "steady_tpot_ms": 43.050256000000005,
          "steady_tpot_p95_ms": 44.141504,
          "ttft_ms": 1651.994542
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 164,
          "gross_j": {
            "core": 5.281297,
            "dram": 2.783379,
            "package-0": 33.277441,
            "psys": 52.980455,
            "uncore": 0.0
          },
          "logits": {
            "l2": 384.67176099469424,
            "sha256": "2a1e5f084e13626a7729e5e9158bdf8240bf1f797678c5c26a29afe86cd5ec20",
            "sum": 17386.329117944464,
            "top1": 72
          },
          "net_j": {
            "core": 2.7604415749329374,
            "dram": 1.978884313210817,
            "package-0": 27.911064759064274,
            "psys": 39.747870015777856,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1598586,
          "npu_utilization": 0.9676702672786434,
          "seconds": 1.651994542,
          "watts": {
            "core": 3.1969215791755325,
            "dram": 1.684859682787015,
            "package-0": 20.143795971451826,
            "psys": 32.0705992986265,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 441.38267318488283,
            "sha256": "d3c8964f9838d4f83cf8587ff10448c8cc4639e4ad36480f4b6e161a894e35c1",
            "sum": -5983.314064975828,
            "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": 60,
          "gross_j": {
            "core": 4.252308,
            "dram": 1.769588,
            "package-0": 10.067967,
            "psys": 17.662552,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.3296073674417026,
            "dram": 1.475121387988555,
            "package-0": 8.10372950235553,
            "psys": 12.819071325181696,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 592436,
          "npu_utilization": 0.9797605609372482,
          "seconds": 0.604674268,
          "token_latencies_s": [
            0.043877501,
            0.043118939,
            0.043086678,
            0.043013834,
            0.042913365,
            0.042845764,
            0.043220444,
            0.043998312,
            0.042664596,
            0.044407432,
            0.042774459,
            0.043160995,
            0.042820934,
            0.042737878
          ],
          "watts": {
            "core": 7.0323945056646595,
            "dram": 2.926514478370361,
            "package-0": 16.6502322536404,
            "psys": 29.210027505255113,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 25.7384000280872,
    "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": 409469,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.026572737842798233,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 20.187692,
            "dram": 11.0795,
            "package-0": 103.328838,
            "psys": 172.386338,
            "uncore": 0.0
          },
          "net_j": {
            "core": 11.103111635187604,
            "dram": 8.180287073099805,
            "package-0": 83.98965823434425,
            "psys": 124.69915985880303,
            "uncore": 0.0
          }
        },
        "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": 17,
          "gross_j": {
            "core": 3.254264,
            "dram": 0.264526,
            "package-0": 4.026052,
            "psys": 5.968063,
            "uncore": 0.0
          },
          "logits": {
            "l2": 479.5725935312274,
            "sha256": "0c2ef8bebe7bf52d4e6a3605f2b28cbeae5302dda9b1e081f94832c5649560d8",
            "sum": 99441.07571207639,
            "top1": 546
          },
          "net_j": {
            "core": 2.992744204549545,
            "dram": 0.18106572542892332,
            "package-0": 3.469330816038266,
            "psys": 4.595281806781639,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 47906,
          "npu_utilization": 0.27952758542711004,
          "seconds": 0.171382012,
          "watts": {
            "core": 18.988363842991877,
            "dram": 1.543487539404077,
            "package-0": 23.491683596292475,
            "psys": 34.82315868715557,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1376.0189132402104,
          "decode_tokens_per_s": 17.851990135218756,
          "end_to_end_ms": 5953.406542000001,
          "handoff_estimate_ms": 124.313991,
          "handoff_plus_first_decode_ms": 171.382012,
          "output_tokens_per_s_e2e": 2.687536939922286,
          "prefill_tokens_per_s": 1599.009900579588,
          "steady_tpot_ms": 47.068021,
          "steady_tpot_p95_ms": 50.8961629,
          "ttft_ms": 5113.164088
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 505,
          "gross_j": {
            "core": 12.421477,
            "dram": 8.858436,
            "package-0": 88.282123,
            "psys": 147.004995,
            "uncore": 0.0
          },
          "logits": {
            "l2": 393.33541703766207,
            "sha256": "73521a985b1e954f610989386d69c028e3b55f95b4a5e007c6f02207bfccaece",
            "sum": 10675.848786275834,
            "top1": 13
          },
          "net_j": {
            "core": 4.619061736031843,
            "dram": 6.368407590767362,
            "package-0": 71.67240572779305,
            "psys": 106.04821420409061,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 5012704,
          "npu_utilization": 0.9803526571275566,
          "seconds": 5.113164088,
          "watts": {
            "core": 2.429313197507539,
            "dram": 1.7324763781373094,
            "package-0": 17.265654197796593,
            "psys": 28.75029873283425,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 530.2405363191889,
            "sha256": "fe7a0eed3c4c1cc5e8a1c731ca976a3a6dc917e8d9768da56d0cca26eb6e651d",
            "sum": 99033.82702467777,
            "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": 66,
          "gross_j": {
            "core": 4.511951,
            "dram": 1.956538,
            "package-0": 11.020663,
            "psys": 19.41328,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.4913056946062158,
            "dram": 1.630813756903521,
            "package-0": 8.847921690512932,
            "psys": 14.055663847930784,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 650694,
          "npu_utilization": 0.9728397123536272,
          "seconds": 0.668860442,
          "token_latencies_s": [
            0.056285127,
            0.047994413,
            0.047450013,
            0.047105563,
            0.047289358,
            0.046911115,
            0.047139601,
            0.047166355,
            0.046876048,
            0.046754606,
            0.046913468,
            0.047030479,
            0.046919979,
            0.046996221
          ],
          "watts": {
            "core": 6.745728580551934,
            "dram": 2.925181214409448,
            "package-0": 16.476774986193607,
            "psys": 29.02441044644706,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.814593,
            "dram": 11.035738,
            "package-0": 99.945911,
            "psys": 166.599305,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.70228147977099,
            "dram": 8.127675074417079,
            "package-0": 80.54769737937518,
            "psys": 118.76655926015694,
            "uncore": 0.0
          }
        },
        "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": 15,
          "gross_j": {
            "core": 2.849968,
            "dram": 0.251098,
            "package-0": 3.583365,
            "psys": 5.322679,
            "uncore": 0.0
          },
          "logits": {
            "l2": 479.5725935312274,
            "sha256": "0c2ef8bebe7bf52d4e6a3605f2b28cbeae5302dda9b1e081f94832c5649560d8",
            "sum": 99441.07571207639,
            "top1": 546
          },
          "net_j": {
            "core": 2.623502138040201,
            "dram": 0.17882468322264283,
            "package-0": 3.1012663405849064,
            "psys": 4.133904453626146,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 47741,
          "npu_utilization": 0.32168299074823653,
          "seconds": 0.148410085,
          "watts": {
            "core": 19.203331094379468,
            "dram": 1.6919200605538363,
            "package-0": 24.14502356763693,
            "psys": 35.864671865122915,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1371.831326560963,
          "decode_tokens_per_s": 18.334684491609448,
          "end_to_end_ms": 5971.579626000001,
          "handoff_estimate_ms": 101.13829499999999,
          "handoff_plus_first_decode_ms": 148.410085,
          "output_tokens_per_s_e2e": 2.6793580596893807,
          "prefill_tokens_per_s": 1586.5075133220696,
          "steady_tpot_ms": 47.27179,
          "steady_tpot_p95_ms": 50.53907135000001,
          "ttft_ms": 5153.458103
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 509,
          "gross_j": {
            "core": 9.585669,
            "dram": 8.824562,
            "package-0": 85.456507,
            "psys": 141.989871,
            "uncore": 0.0
          },
          "logits": {
            "l2": 393.33541703766207,
            "sha256": "73521a985b1e954f610989386d69c028e3b55f95b4a5e007c6f02207bfccaece",
            "sum": 10675.848786275834,
            "top1": 13
          },
          "net_j": {
            "core": 1.7217672215621826,
            "dram": 6.314911055611975,
            "package-0": 68.71589774283157,
            "psys": 100.71033249255635,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 5035934,
          "npu_utilization": 0.9771950987761819,
          "seconds": 5.153458103,
          "watts": {
            "core": 1.8600459746475597,
            "dram": 1.7123573770519116,
            "package-0": 16.582361841702546,
            "psys": 27.552347988885938,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 530.2405363191889,
            "sha256": "fe7a0eed3c4c1cc5e8a1c731ca976a3a6dc917e8d9768da56d0cca26eb6e651d",
            "sum": 99033.82702467777,
            "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": 66,
          "gross_j": {
            "core": 4.378956,
            "dram": 1.960078,
            "package-0": 10.906039,
            "psys": 19.286755,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.357012120168607,
            "dram": 1.6339393355824614,
            "package-0": 8.730533295958695,
            "psys": 13.922322313974442,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 652629,
          "npu_utilization": 0.9744928382125079,
          "seconds": 0.669711438,
          "token_latencies_s": [
            0.052577784,
            0.047581958,
            0.047213935,
            0.047473785,
            0.047158253,
            0.047083428,
            0.047261864,
            0.047235505,
            0.047223046,
            0.049441303,
            0.04754857,
            0.04726505,
            0.047315954,
            0.04727853
          ],
          "watts": {
            "core": 6.538571318233928,
            "dram": 2.9267500729172258,
            "package-0": 16.28468379242464,
            "psys": 28.798604750722504,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.303413,
            "dram": 11.120089,
            "package-0": 104.836767,
            "psys": 173.978315,
            "uncore": 0.0
          },
          "net_j": {
            "core": 12.240999615609848,
            "dram": 8.227950345505326,
            "package-0": 85.54477611946963,
            "psys": 126.40749674575414,
            "uncore": 0.0
          }
        },
        "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": 15,
          "gross_j": {
            "core": 2.893181,
            "dram": 0.253051,
            "package-0": 3.63323,
            "psys": 5.374987,
            "uncore": 0.0
          },
          "logits": {
            "l2": 479.5725935312274,
            "sha256": "0c2ef8bebe7bf52d4e6a3605f2b28cbeae5302dda9b1e081f94832c5649560d8",
            "sum": 99441.07571207639,
            "top1": 546
          },
          "net_j": {
            "core": 2.6607408374348425,
            "dram": 0.17887107126604324,
            "package-0": 3.138413298722609,
            "psys": 4.15485189256597,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 47768,
          "npu_utilization": 0.31359217585683064,
          "seconds": 0.152325229,
          "watts": {
            "core": 18.99344592483757,
            "dram": 1.6612546828995742,
            "package-0": 23.85179411087575,
            "psys": 35.28625583093658,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1379.3847036777745,
          "decode_tokens_per_s": 18.403970998811467,
          "end_to_end_ms": 5938.879833999999,
          "handoff_estimate_ms": 105.13715850000001,
          "handoff_plus_first_decode_ms": 152.325229,
          "output_tokens_per_s_e2e": 2.6941107493706533,
          "prefill_tokens_per_s": 1595.6787579688396,
          "steady_tpot_ms": 47.188070499999995,
          "steady_tpot_p95_ms": 47.919905050000004,
          "ttft_ms": 5123.838341
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 505,
          "gross_j": {
            "core": 14.846276,
            "dram": 8.915383,
            "package-0": 91.142955,
            "psys": 150.666728,
            "uncore": 0.0
          },
          "logits": {
            "l2": 393.33541703766207,
            "sha256": "73521a985b1e954f610989386d69c028e3b55f95b4a5e007c6f02207bfccaece",
            "sum": 10675.848786275834,
            "top1": 13
          },
          "net_j": {
            "core": 7.027572396149672,
            "dram": 6.42015640178694,
            "package-0": 74.49856324417182,
            "psys": 109.62444573518468,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 5009495,
          "npu_utilization": 0.9776840459455862,
          "seconds": 5.123838341,
          "watts": {
            "core": 2.897491101778693,
            "dram": 1.7399813199922345,
            "package-0": 17.788023144815295,
            "psys": 29.40505105213662,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 530.2405363191889,
            "sha256": "fe7a0eed3c4c1cc5e8a1c731ca976a3a6dc917e8d9768da56d0cca26eb6e651d",
            "sum": 99033.82702467777,
            "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": 66,
          "gross_j": {
            "core": 3.563956,
            "dram": 1.951655,
            "package-0": 10.060582,
            "psys": 17.9366,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.5526863820253327,
            "dram": 1.6289228724523426,
            "package-0": 7.907799576575208,
            "psys": 12.6281991180035,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 650144,
          "npu_utilization": 0.9810291904953158,
          "seconds": 0.662716264,
          "token_latencies_s": [
            0.048366847,
            0.047633622,
            0.047679244,
            0.047137533,
            0.047037577,
            0.047077295,
            0.047227845,
            0.047531262,
            0.047332575,
            0.047349869,
            0.047137763,
            0.047148296,
            0.047035432,
            0.04698602
          ],
          "watts": {
            "core": 5.377800717442479,
            "dram": 2.9449330068048547,
            "package-0": 15.18082857854836,
            "psys": 27.065278120290703,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 14.877661527134478,
    "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": 409687,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "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": 0.04378183372318745,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 12.625821,
            "dram": 6.100204,
            "package-0": 42.698865,
            "psys": 71.28888,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.056031138021263,
            "dram": 4.96095700139857,
            "package-0": 35.09952577968502,
            "psys": 52.55018231241183,
            "uncore": 0.0
          }
        },
        "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": 21,
          "gross_j": {
            "core": 3.176627,
            "dram": 0.395995,
            "package-0": 4.39751,
            "psys": 6.947797,
            "uncore": 0.0
          },
          "logits": {
            "l2": 530.8437238572551,
            "sha256": "274c0813f3a62c60b62c0cc581677f1b44f808364130fe3aac5dcc63f5f49df1",
            "sum": -124148.91462222487,
            "top1": 546
          },
          "net_j": {
            "core": 2.853316875641328,
            "dram": 0.292815232273026,
            "package-0": 3.709250073405361,
            "psys": 5.250663299156468,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 98585,
          "npu_utilization": 0.4652976583952569,
          "seconds": 0.211875126,
          "watts": {
            "core": 14.992920877366227,
            "dram": 1.869001838375308,
            "package-0": 20.75519709661435,
            "psys": 32.79194274083877,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 437.7202469945753,
          "decode_tokens_per_s": 9.549546023272141,
          "end_to_end_ms": 2339.393727,
          "handoff_estimate_ms": 114.9072025,
          "handoff_plus_first_decode_ms": 211.875126,
          "output_tokens_per_s_e2e": 6.839378859290239,
          "prefill_tokens_per_s": 1311.4098393440984,
          "steady_tpot_ms": 96.9679235,
          "steady_tpot_p95_ms": 97.5769649,
          "ttft_ms": 768.63843
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 76,
          "gross_j": {
            "core": 2.516473,
            "dram": 1.489071,
            "package-0": 17.759964,
            "psys": 27.334037,
            "uncore": 0.0
          },
          "logits": {
            "l2": 868.4674711122068,
            "sha256": "1773fbd9c1c18029e107f5353f56d0ed01120e9f0f0ea934e28c472a840595d4",
            "sum": -269072.11017668527,
            "top1": 198
          },
          "net_j": {
            "core": 1.3435718148455214,
            "dram": 1.114756486679423,
            "package-0": 15.263101629239127,
            "psys": 21.177192598564872,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 760854,
          "npu_utilization": 0.9898724423653915,
          "seconds": 0.76863843,
          "watts": {
            "core": 3.2739359649243664,
            "dram": 1.9372840881765436,
            "package-0": 23.105745571425565,
            "psys": 35.56163201467822,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 520.1749807146045,
            "sha256": "631550d11dadb58c7812a3211e73e0117bf885be1946c6a0b54350ad0baa4cd2",
            "sum": -88494.78699321579,
            "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": 134,
          "gross_j": {
            "core": 6.932721,
            "dram": 4.215138,
            "package-0": 20.541391,
            "psys": 37.007046,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.859142447534413,
            "dram": 3.5533852824461203,
            "package-0": 16.12717407704053,
            "psys": 26.122326414690487,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1347943,
          "npu_utilization": 0.9919513351998127,
          "seconds": 1.358880171,
          "token_latencies_s": [
            0.097993169,
            0.097352855,
            0.097348436,
            0.097027084,
            0.09702352,
            0.096966228,
            0.096902958,
            0.096771809,
            0.09681007,
            0.096853403,
            0.096969619,
            0.09691046,
            0.096922908,
            0.096996998
          ],
          "watts": {
            "core": 5.101789803068662,
            "dram": 3.1019203090571845,
            "package-0": 15.116410878880947,
            "psys": 27.23348738893328,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.053617,
            "dram": 6.079024,
            "package-0": 42.9568,
            "psys": 70.633853,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.64327241591715,
            "dram": 4.990661664448,
            "package-0": 35.696886572422464,
            "psys": 52.7321224661446,
            "uncore": 0.0
          }
        },
        "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": 2.419733,
            "dram": 0.322448,
            "package-0": 3.423209,
            "psys": 5.009082,
            "uncore": 0.0
          },
          "logits": {
            "l2": 530.8437238572551,
            "sha256": "274c0813f3a62c60b62c0cc581677f1b44f808364130fe3aac5dcc63f5f49df1",
            "sum": -124148.91462222487,
            "top1": 546
          },
          "net_j": {
            "core": 2.252264667306225,
            "dram": 0.26900289461320426,
            "package-0": 3.0667037943700164,
            "psys": 4.129999874741901,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 96679,
          "npu_utilization": 0.8809246932898385,
          "seconds": 0.109747179,
          "watts": {
            "core": 22.048247818743477,
            "dram": 2.93809829954718,
            "package-0": 31.19177213657583,
            "psys": 45.64201144523268,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 458.1851662137208,
          "decode_tokens_per_s": 10.202238934793973,
          "end_to_end_ms": 2234.904304,
          "handoff_estimate_ms": 12.603467499999994,
          "handoff_plus_first_decode_ms": 109.747179,
          "output_tokens_per_s_e2e": 7.159143222089387,
          "prefill_tokens_per_s": 1318.2694939817445,
          "steady_tpot_ms": 97.14371150000001,
          "steady_tpot_p95_ms": 97.51050240000001,
          "ttft_ms": 764.6387970000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 76,
          "gross_j": {
            "core": 2.990349,
            "dram": 1.542476,
            "package-0": 18.239273,
            "psys": 27.717397,
            "uncore": 0.0
          },
          "logits": {
            "l2": 868.4674711122068,
            "sha256": "1773fbd9c1c18029e107f5353f56d0ed01120e9f0f0ea934e28c472a840595d4",
            "sum": -269072.11017668527,
            "top1": 198
          },
          "net_j": {
            "core": 1.8235510411516573,
            "dram": 1.1701092434157807,
            "package-0": 15.755403127070071,
            "psys": 21.592589922074502,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 759652,
          "npu_utilization": 0.9934782317879169,
          "seconds": 0.764638797,
          "watts": {
            "core": 3.9107994673202544,
            "dram": 2.0172609682529616,
            "package-0": 23.85344959157232,
            "psys": 36.24900686277889,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 520.1749807146045,
            "sha256": "631550d11dadb58c7812a3211e73e0117bf885be1946c6a0b54350ad0baa4cd2",
            "sum": -88494.78699321579,
            "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": 7.643535,
            "dram": 4.2141,
            "package-0": 21.294318,
            "psys": 37.907374,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.567456707459267,
            "dram": 3.5515495264190147,
            "package-0": 16.87477965098238,
            "psys": 27.009532669328202,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1348127,
          "npu_utilization": 0.9908922006084141,
          "seconds": 1.360518328,
          "token_latencies_s": [
            0.097487604,
            0.097553028,
            0.097196225,
            0.097057119,
            0.097248958,
            0.096993922,
            0.097084041,
            0.097016508,
            0.097074958,
            0.097104771,
            0.097215076,
            0.096961535,
            0.097312637,
            0.097182652
          ],
          "watts": {
            "core": 5.618105131473099,
            "dram": 3.097422440603829,
            "package-0": 15.651621563454603,
            "psys": 27.86245008233362,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.000089,
            "dram": 6.08464,
            "package-0": 42.913464,
            "psys": 70.08771200000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.599591088274787,
            "dram": 4.999420088053695,
            "package-0": 35.674512087337945,
            "psys": 52.23766905103686,
            "uncore": 0.0
          }
        },
        "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": 2.392633,
            "dram": 0.317992,
            "package-0": 3.406607,
            "psys": 4.953052,
            "uncore": 0.0
          },
          "logits": {
            "l2": 530.8437238572551,
            "sha256": "274c0813f3a62c60b62c0cc581677f1b44f808364130fe3aac5dcc63f5f49df1",
            "sum": -124148.91462222487,
            "top1": 546
          },
          "net_j": {
            "core": 2.2288425602185056,
            "dram": 0.26572064114269045,
            "package-0": 3.0579312624723536,
            "psys": 4.093276031707773,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 96608,
          "npu_utilization": 0.9000442736667493,
          "seconds": 0.107336942,
          "watts": {
            "core": 22.290862357528315,
            "dram": 2.962558780554788,
            "package-0": 31.73750748367696,
            "psys": 46.14489576198285,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 459.5119128631688,
          "decode_tokens_per_s": 10.213703232918563,
          "end_to_end_ms": 2228.451475,
          "handoff_estimate_ms": 10.34657800000001,
          "handoff_plus_first_decode_ms": 107.33694200000001,
          "output_tokens_per_s_e2e": 7.179873638487012,
          "prefill_tokens_per_s": 1326.6016075355562,
          "steady_tpot_ms": 96.990364,
          "steady_tpot_p95_ms": 98.19586815000001,
          "ttft_ms": 759.836257
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 76,
          "gross_j": {
            "core": 3.143608,
            "dram": 1.542171,
            "package-0": 18.388259,
            "psys": 27.916249,
            "uncore": 0.0
          },
          "logits": {
            "l2": 868.4674711122068,
            "sha256": "1773fbd9c1c18029e107f5353f56d0ed01120e9f0f0ea934e28c472a840595d4",
            "sum": -269072.11017668527,
            "top1": 198
          },
          "net_j": {
            "core": 1.9841384606502135,
            "dram": 1.1721430029061208,
            "package-0": 15.919989806065622,
            "psys": 21.829910583486175,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 753285,
          "npu_utilization": 0.9913780673932752,
          "seconds": 0.759836257,
          "watts": {
            "core": 4.1372176847833675,
            "dram": 2.02960965049079,
            "package-0": 24.20029161625016,
            "psys": 36.73982222198697,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 520.1749807146045,
            "sha256": "631550d11dadb58c7812a3211e73e0117bf885be1946c6a0b54350ad0baa4cd2",
            "sum": -88494.78699321579,
            "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": 7.463848,
            "dram": 4.224477,
            "package-0": 21.118598,
            "psys": 37.218411,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.3866100674060675,
            "dram": 3.561556444004883,
            "package-0": 16.696591018799968,
            "psys": 26.314482435842912,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1350546,
          "npu_utilization": 0.9921160308004504,
          "seconds": 1.361278276,
          "token_latencies_s": [
            0.097855952,
            0.097332537,
            0.097250902,
            0.096996809,
            0.096983919,
            0.096845612,
            0.096913275,
            0.096857339,
            0.097707471,
            0.096925419,
            0.096755145,
            0.096981384,
            0.097019536,
            0.098827141
          ],
          "watts": {
            "core": 5.482970037494376,
            "dram": 3.103316253906046,
            "package-0": 15.513799325480457,
            "psys": 27.34078083532129,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 32.750238213688135,
    "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": 409811,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.04384459741413593,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 13.805689999999998,
            "dram": 8.016215,
            "package-0": 62.817343,
            "psys": 103.568033,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.931583572579628,
            "dram": 6.460714162977434,
            "package-0": 52.44138543107081,
            "psys": 77.9826595265222,
            "uncore": 0.0
          }
        },
        "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": 3.318778,
            "dram": 0.416197,
            "package-0": 4.59093,
            "psys": 7.060834,
            "uncore": 0.0
          },
          "logits": {
            "l2": 518.0133203759266,
            "sha256": "b5b42791fa53b84d4c0febbc65f92575aa94c4de03d3a76de6c941f2c254cd68",
            "sum": -135962.72018502932,
            "top1": 319
          },
          "net_j": {
            "core": 2.9977560966309924,
            "dram": 0.31374748503170447,
            "package-0": 3.907541219378606,
            "psys": 5.375711729247749,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 101218,
          "npu_utilization": 0.48112997751678255,
          "seconds": 0.210375584,
          "watts": {
            "core": 15.775490372494938,
            "dram": 1.9783522026966778,
            "package-0": 21.822541916271046,
            "psys": 33.56299179661457,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 641.1715966288639,
          "decode_tokens_per_s": 9.262187951654772,
          "end_to_end_ms": 3194.152721,
          "handoff_estimate_ms": 109.8744865,
          "handoff_plus_first_decode_ms": 210.375584,
          "output_tokens_per_s_e2e": 5.009153098662999,
          "prefill_tokens_per_s": 1290.4332335741442,
          "steady_tpot_ms": 100.5010975,
          "steady_tpot_p95_ms": 101.3804163,
          "ttft_ms": 1574.664963
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 156,
          "gross_j": {
            "core": 4.209401,
            "dram": 3.225822,
            "package-0": 37.782068,
            "psys": 58.820711,
            "uncore": 0.0
          },
          "logits": {
            "l2": 805.6627491508196,
            "sha256": "29e78d35151f46938d24fab3610c72064af8b845ed8c88ce1e8e01295747b58c",
            "sum": -252043.32194374222,
            "top1": 319
          },
          "net_j": {
            "core": 1.8065463817057608,
            "dram": 2.4589855704601584,
            "package-0": 32.66689090344205,
            "psys": 46.207541031927235,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1535265,
          "npu_utilization": 0.9749788279248073,
          "seconds": 1.574664963,
          "watts": {
            "core": 2.673204204645785,
            "dram": 2.0485767295249078,
            "package-0": 23.993718592695963,
            "psys": 37.3544292799509,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 543.8257039866751,
            "sha256": "789c4fe7e50b20ef2d86f9de8d79eba5278419cb5116623fcb97983782b06aec",
            "sum": -117196.24898801185,
            "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": 140,
          "gross_j": {
            "core": 6.277511,
            "dram": 4.374196,
            "package-0": 20.444345,
            "psys": 37.686488,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.127281094242875,
            "dram": 3.6879811074855704,
            "package-0": 15.866953308250151,
            "psys": 26.39940676534721,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1398688,
          "npu_utilization": 0.9926023107369731,
          "seconds": 1.409112174,
          "token_latencies_s": [
            0.102201534,
            0.100752478,
            0.10086571,
            0.100707824,
            0.100495026,
            0.100920141,
            0.100430559,
            0.100507169,
            0.100217411,
            0.100224728,
            0.100303294,
            0.100251548,
            0.100938276,
            0.100267664
          ],
          "watts": {
            "core": 4.454940575937426,
            "dram": 3.1042212825279303,
            "package-0": 14.508671046369086,
            "psys": 26.744845935877915,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.872947,
            "dram": 7.976053,
            "package-0": 63.652914,
            "psys": 102.079144,
            "uncore": 0.0
          },
          "net_j": {
            "core": 10.051472847622168,
            "dram": 6.437348995197283,
            "package-0": 53.3889995841553,
            "psys": 76.77005017567933,
            "uncore": 0.0
          }
        },
        "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": 2.762261,
            "dram": 0.394835,
            "package-0": 3.963674,
            "psys": 6.046371,
            "uncore": 0.0
          },
          "logits": {
            "l2": 518.0133203759266,
            "sha256": "b5b42791fa53b84d4c0febbc65f92575aa94c4de03d3a76de6c941f2c254cd68",
            "sum": -135962.72018502932,
            "top1": 319
          },
          "net_j": {
            "core": 2.4938888101910712,
            "dram": 0.3091878825031615,
            "package-0": 3.392365495435437,
            "psys": 4.637619917436837,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 100883,
          "npu_utilization": 0.5736140885347017,
          "seconds": 0.175872598,
          "watts": {
            "core": 15.706033978073151,
            "dram": 2.2450057853810748,
            "package-0": 22.537189107765386,
            "psys": 34.37926697369877,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 648.170767163998,
          "decode_tokens_per_s": 9.476649174741327,
          "end_to_end_ms": 3159.661163,
          "handoff_estimate_ms": 75.40130049999999,
          "handoff_plus_first_decode_ms": 175.87259799999998,
          "output_tokens_per_s_e2e": 5.063834118468734,
          "prefill_tokens_per_s": 1288.6669790821597,
          "steady_tpot_ms": 100.4712975,
          "steady_tpot_p95_ms": 100.77077915,
          "ttft_ms": 1576.823208
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 156,
          "gross_j": {
            "core": 4.733874,
            "dram": 3.217032,
            "package-0": 38.154017,
            "psys": 58.092319,
            "uncore": 0.0
          },
          "logits": {
            "l2": 805.6627491508196,
            "sha256": "29e78d35151f46938d24fab3610c72064af8b845ed8c88ce1e8e01295747b58c",
            "sum": -252043.32194374222,
            "top1": 319
          },
          "net_j": {
            "core": 2.3277260151246058,
            "dram": 2.4491445399687457,
            "package-0": 33.03182901182246,
            "psys": 45.46186134746268,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1536304,
          "npu_utilization": 0.9743032650747235,
          "seconds": 1.576823208,
          "watts": {
            "core": 3.002159009318691,
            "dram": 2.0401982820131095,
            "package-0": 24.196762710255598,
            "psys": 36.84136478031848,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 543.8257039866751,
            "sha256": "789c4fe7e50b20ef2d86f9de8d79eba5278419cb5116623fcb97983782b06aec",
            "sum": -117196.24898801185,
            "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": 7.376812,
            "dram": 4.364186,
            "package-0": 21.535223,
            "psys": 37.940454,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.229858022306491,
            "dram": 3.6790165727253754,
            "package-0": 16.964805076897406,
            "psys": 26.670568910779807,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1396469,
          "npu_utilization": 0.9925397189434849,
          "seconds": 1.406965357,
          "token_latencies_s": [
            0.100761263,
            0.100536781,
            0.100481018,
            0.100359699,
            0.100771929,
            0.100412653,
            0.100448994,
            0.100331518,
            0.100461577,
            0.100310847,
            0.100209227,
            0.100551201,
            0.100525837,
            0.10077016
          ],
          "watts": {
            "core": 5.243065839040414,
            "dram": 3.1018432531313564,
            "package-0": 15.306150142828285,
            "psys": 26.96616075956446,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 15.432638,
            "dram": 7.9350380000000005,
            "package-0": 64.16756099999999,
            "psys": 101.470932,
            "uncore": 0.0
          },
          "net_j": {
            "core": 10.588772809411378,
            "dram": 6.389188217976919,
            "package-0": 53.85598072665732,
            "psys": 76.04430215474652,
            "uncore": 0.0
          }
        },
        "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": 2.758843,
            "dram": 0.392028,
            "package-0": 3.971426,
            "psys": 5.970383,
            "uncore": 0.0
          },
          "logits": {
            "l2": 518.0133203759266,
            "sha256": "b5b42791fa53b84d4c0febbc65f92575aa94c4de03d3a76de6c941f2c254cd68",
            "sum": -135962.72018502932,
            "top1": 319
          },
          "net_j": {
            "core": 2.486692702262956,
            "dram": 0.3051751537953345,
            "package-0": 3.3920746904338017,
            "psys": 4.541799707409552,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 100967,
          "npu_utilization": 0.5661219180699739,
          "seconds": 0.178348509,
          "watts": {
            "core": 15.468831309377531,
            "dram": 2.198100798252258,
            "package-0": 22.26778357872339,
            "psys": 33.47593446940451,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 645.1745614803872,
          "decode_tokens_per_s": 9.428149969795225,
          "end_to_end_ms": 3174.334703,
          "handoff_estimate_ms": 77.75900549999999,
          "handoff_plus_first_decode_ms": 178.34850899999998,
          "output_tokens_per_s_e2e": 5.040426261565525,
          "prefill_tokens_per_s": 1283.351268725201,
          "steady_tpot_ms": 100.58950349999999,
          "steady_tpot_p95_ms": 102.60561775,
          "ttft_ms": 1583.354495
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 157,
          "gross_j": {
            "core": 5.208482,
            "dram": 3.173087,
            "package-0": 38.567406,
            "psys": 57.646703,
            "uncore": 0.0
          },
          "logits": {
            "l2": 805.6627491508196,
            "sha256": "29e78d35151f46938d24fab3610c72064af8b845ed8c88ce1e8e01295747b58c",
            "sum": -252043.32194374222,
            "top1": 319
          },
          "net_j": {
            "core": 2.7923676200500265,
            "dram": 2.4020189085829826,
            "package-0": 33.42400163217444,
            "psys": 44.96392930882782,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1537599,
          "npu_utilization": 0.9711021788585631,
          "seconds": 1.583354495,
          "watts": {
            "core": 3.2895236135986083,
            "dram": 2.0040281630046466,
            "package-0": 24.3580361326476,
            "psys": 36.40795739806833,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 543.8257039866751,
            "sha256": "789c4fe7e50b20ef2d86f9de8d79eba5278419cb5116623fcb97983782b06aec",
            "sum": -117196.24898801185,
            "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": 140,
          "gross_j": {
            "core": 7.465313,
            "dram": 4.369923,
            "package-0": 21.628729,
            "psys": 37.853846,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.309712487098395,
            "dram": 3.681994155598602,
            "package-0": 17.039904404049082,
            "psys": 26.53857313850914,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1400999,
          "npu_utilization": 0.9917652286804588,
          "seconds": 1.412631699,
          "token_latencies_s": [
            0.101521256,
            0.100695076,
            0.100341163,
            0.100861898,
            0.10039921,
            0.100323068,
            0.100266442,
            0.102911108,
            0.100601021,
            0.100703905,
            0.100577986,
            0.100543907,
            0.100402046,
            0.102441123
          ],
          "watts": {
            "core": 5.284684610493085,
            "dram": 3.0934623675041855,
            "package-0": 15.310946947679955,
            "psys": 26.79668453341142,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 31.873993958346546,
    "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": 409987,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.04428494721651077,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 18.219131,
            "dram": 12.717863,
            "package-0": 110.94386899999999,
            "psys": 182.816183,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.644810345639879,
            "dram": 9.981492107537589,
            "package-0": 92.69092589815648,
            "psys": 137.8074825356435,
            "uncore": 0.0
          }
        },
        "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": 28,
          "gross_j": {
            "core": 4.115285,
            "dram": 0.525572,
            "package-0": 5.530137,
            "psys": 8.811318,
            "uncore": 0.0
          },
          "logits": {
            "l2": 609.263242568076,
            "sha256": "1f93e2afdd0c63c174d4669ef6edf80514b7042aecb99e298ed65730a9a55a74",
            "sum": -156409.51710008364,
            "top1": 437
          },
          "net_j": {
            "core": 3.68111987681363,
            "dram": 0.387014456667933,
            "package-0": 4.605889866054074,
            "psys": 6.532279371834293,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 129805,
          "npu_utilization": 0.4562215745015789,
          "seconds": 0.284521836,
          "watts": {
            "core": 14.46386350466261,
            "dram": 1.847211473779468,
            "package-0": 19.436599586683393,
            "psys": 30.96886384495284,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 728.9530509029807,
          "decode_tokens_per_s": 8.336893061617248,
          "end_to_end_ms": 5619.017569,
          "handoff_estimate_ms": 176.70001949999997,
          "handoff_plus_first_decode_ms": 284.52183599999995,
          "output_tokens_per_s_e2e": 2.8474728550897686,
          "prefill_tokens_per_s": 1068.1226247560955,
          "steady_tpot_ms": 107.8218165,
          "steady_tpot_p95_ms": 110.00807129999998,
          "ttft_ms": 3819.786142
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 378,
          "gross_j": {
            "core": 8.283426,
            "dram": 7.514996,
            "package-0": 84.643277,
            "psys": 135.13442,
            "uncore": 0.0
          },
          "logits": {
            "l2": 657.7937454973483,
            "sha256": "01330236be5a5d092e077e2f3012d7ef6a5af180540b794e147c26e69be8d496",
            "sum": -175698.30940290913,
            "top1": 72
          },
          "net_j": {
            "core": 2.4546363921365826,
            "dram": 5.654821781633876,
            "package-0": 72.23499774922972,
            "psys": 104.53768165704183,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3706905,
          "npu_utilization": 0.970448308412131,
          "seconds": 3.819786142,
          "watts": {
            "core": 2.168557529679629,
            "dram": 1.9673865815077352,
            "package-0": 22.159166469901287,
            "psys": 35.37748318266976,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 474.8007585329114,
            "sha256": "56e5d3e7ac8fb688d7b620d385293c7321b36fd5df528c679a4978f4fa7ebb39",
            "sum": -66282.34790089168,
            "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": 149,
          "gross_j": {
            "core": 5.82042,
            "dram": 4.677295,
            "package-0": 20.770455,
            "psys": 38.870445,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.5090540766896665,
            "dram": 3.9396558692357795,
            "package-0": 15.850038282872688,
            "psys": 26.737521506767372,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1496238,
          "npu_utilization": 0.9878051930813978,
          "seconds": 1.514709591,
          "token_latencies_s": [
            0.112792891,
            0.107809922,
            0.107868182,
            0.107524315,
            0.107588909,
            0.107484573,
            0.107572654,
            0.107596105,
            0.107842735,
            0.108508553,
            0.107833711,
            0.108426957,
            0.108072982,
            0.107749612
          ],
          "watts": {
            "core": 3.8425979703194475,
            "dram": 3.087915352085468,
            "package-0": 13.712499824000917,
            "psys": 25.661978527737467,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.34748,
            "dram": 12.823758999999999,
            "package-0": 114.036085,
            "psys": 183.34462299999998,
            "uncore": 0.0
          },
          "net_j": {
            "core": 12.83386607851687,
            "dram": 10.106761786633276,
            "package-0": 95.91237389049387,
            "psys": 138.65458698340248,
            "uncore": 0.0
          }
        },
        "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": 3.707388,
            "dram": 0.512022,
            "package-0": 4.968127,
            "psys": 7.922587,
            "uncore": 0.0
          },
          "logits": {
            "l2": 609.263242568076,
            "sha256": "1f93e2afdd0c63c174d4669ef6edf80514b7042aecb99e298ed65730a9a55a74",
            "sum": -156409.51710008364,
            "top1": 437
          },
          "net_j": {
            "core": 3.2999981823430886,
            "dram": 0.38200940963168073,
            "package-0": 4.10087891566568,
            "psys": 5.784098481282815,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 130458,
          "npu_utilization": 0.4886522210025914,
          "seconds": 0.26697515,
          "watts": {
            "core": 13.886640760385376,
            "dram": 1.9178638910775028,
            "package-0": 18.60894918497096,
            "psys": 29.675372408255974,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 734.1508856355811,
          "decode_tokens_per_s": 8.466271671821122,
          "end_to_end_ms": 5579.234569,
          "handoff_estimate_ms": 159.79702400000002,
          "handoff_plus_first_decode_ms": 266.97515,
          "output_tokens_per_s_e2e": 2.8677768970139885,
          "prefill_tokens_per_s": 1071.5697301983532,
          "steady_tpot_ms": 107.17812599999999,
          "steady_tpot_p95_ms": 108.85781275000001,
          "ttft_ms": 3807.4983690000004
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 376,
          "gross_j": {
            "core": 10.478367,
            "dram": 7.644756,
            "package-0": 86.89376,
            "psys": 135.780353,
            "uncore": 0.0
          },
          "logits": {
            "l2": 657.7937454973483,
            "sha256": "01330236be5a5d092e077e2f3012d7ef6a5af180540b794e147c26e69be8d496",
            "sum": -175698.30940290913,
            "top1": 72
          },
          "net_j": {
            "core": 4.668327877347957,
            "dram": 5.790565728819927,
            "package-0": 74.52539662752395,
            "psys": 105.28204052731888,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3701084,
          "npu_utilization": 0.9720513684611377,
          "seconds": 3.807498369,
          "watts": {
            "core": 2.7520345340954235,
            "dram": 2.0078159618510396,
            "package-0": 22.821745823313837,
            "psys": 35.661302997658616,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 474.8007585329114,
            "sha256": "56e5d3e7ac8fb688d7b620d385293c7321b36fd5df528c679a4978f4fa7ebb39",
            "sum": -66282.34790089168,
            "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": 149,
          "gross_j": {
            "core": 7.161725,
            "dram": 4.666981,
            "package-0": 22.174198,
            "psys": 39.641683,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.865540018825825,
            "dram": 3.934186648181668,
            "package-0": 17.286098347304236,
            "psys": 27.58844797480078,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1495318,
          "npu_utilization": 0.9937245518150539,
          "seconds": 1.50476105,
          "token_latencies_s": [
            0.11150901,
            0.107430245,
            0.10722571,
            0.107223819,
            0.107261423,
            0.107118572,
            0.107134035,
            0.107060124,
            0.107079946,
            0.107060764,
            0.107010555,
            0.107189788,
            0.107262411,
            0.107166464
          ],
          "watts": {
            "core": 4.759376912367582,
            "dram": 3.10147647694629,
            "package-0": 14.73602602885023,
            "psys": 26.344171388540396,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 26.849235,
            "dram": 12.930997,
            "package-0": 119.537415,
            "psys": 187.20520900000002,
            "uncore": 0.0
          },
          "net_j": {
            "core": 18.319691259784168,
            "dram": 10.208916014511901,
            "package-0": 101.37979262373914,
            "psys": 142.4315534794565,
            "uncore": 0.0
          }
        },
        "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": 26,
          "gross_j": {
            "core": 3.89385,
            "dram": 0.506224,
            "package-0": 5.164355,
            "psys": 8.112955,
            "uncore": 0.0
          },
          "logits": {
            "l2": 609.263242568076,
            "sha256": "1f93e2afdd0c63c174d4669ef6edf80514b7042aecb99e298ed65730a9a55a74",
            "sum": -156409.51710008364,
            "top1": 437
          },
          "net_j": {
            "core": 3.4984456366897096,
            "dram": 0.38003639469220024,
            "package-0": 4.322621452042083,
            "psys": 6.037381054502323,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 129917,
          "npu_utilization": 0.5013763643153943,
          "seconds": 0.259120711,
          "watts": {
            "core": 15.02716623836371,
            "dram": 1.9536223023099069,
            "package-0": 19.93030576394181,
            "psys": 31.30955826992926,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 732.7797817539648,
          "decode_tokens_per_s": 8.494204063666638,
          "end_to_end_ms": 5589.673872,
          "handoff_estimate_ms": 151.9584165,
          "handoff_plus_first_decode_ms": 259.12071100000003,
          "output_tokens_per_s_e2e": 2.862421022476425,
          "prefill_tokens_per_s": 1067.011496886735,
          "steady_tpot_ms": 107.1622945,
          "steady_tpot_p95_ms": 109.60956010000001,
          "ttft_ms": 3823.76386
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 378,
          "gross_j": {
            "core": 14.870262,
            "dram": 7.751933,
            "package-0": 91.300792,
            "psys": 138.592541,
            "uncore": 0.0
          },
          "logits": {
            "l2": 657.7937454973483,
            "sha256": "01330236be5a5d092e077e2f3012d7ef6a5af180540b794e147c26e69be8d496",
            "sum": -175698.30940290913,
            "top1": 72
          },
          "net_j": {
            "core": 9.035402606949793,
            "dram": 5.889821697148918,
            "package-0": 78.87959144057788,
            "psys": 107.96394087412423,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3709788,
          "npu_utilization": 0.9701927566206978,
          "seconds": 3.82376386,
          "watts": {
            "core": 3.8889069891465526,
            "dram": 2.027304322082274,
            "package-0": 23.87720459285893,
            "psys": 36.24505750729074,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 474.8007585329114,
            "sha256": "56e5d3e7ac8fb688d7b620d385293c7321b36fd5df528c679a4978f4fa7ebb39",
            "sum": -66282.34790089168,
            "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": 149,
          "gross_j": {
            "core": 8.085123,
            "dram": 4.67284,
            "package-0": 23.072268,
            "psys": 40.499713,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.785843016144667,
            "dram": 3.9390579226707834,
            "package-0": 18.177579731119174,
            "psys": 28.430231550829944,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1495355,
          "npu_utilization": 0.9924114798317114,
          "seconds": 1.506789301,
          "token_latencies_s": [
            0.113447554,
            0.107542948,
            0.107317305,
            0.10721283,
            0.107154704,
            0.107069677,
            0.107135166,
            0.107200964,
            0.107249605,
            0.107071279,
            0.107074998,
            0.107169885,
            0.106999901,
            0.107105706
          ],
          "watts": {
            "core": 5.365795333583935,
            "dram": 3.101190058157972,
            "package-0": 15.312205883521868,
            "psys": 26.87815275375386,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 96.2730981754139,
    "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": 410177,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.04408681485801935,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 44.705817,
            "dram": 26.179865,
            "package-0": 249.032627,
            "psys": 393.142243,
            "uncore": 0.0
          },
          "net_j": {
            "core": 26.42580749283173,
            "dram": 20.346063356027738,
            "package-0": 210.11829350515575,
            "psys": 297.1860144473222,
            "uncore": 0.0
          }
        },
        "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": 5.891281,
            "dram": 0.733152,
            "package-0": 7.692608,
            "psys": 12.431914,
            "uncore": 0.0
          },
          "logits": {
            "l2": 539.8529262282318,
            "sha256": "21b75f19253b392d43acf6a0728dac2329dc19092f92875416d390c56bfec5b5",
            "sum": -136239.45887422375,
            "top1": 546
          },
          "net_j": {
            "core": 5.240875637733782,
            "dram": 0.5255845000953016,
            "package-0": 6.308030428301388,
            "psys": 9.017777614073562,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 149187,
          "npu_utilization": 0.3500146308072364,
          "seconds": 0.426230754,
          "watts": {
            "core": 13.821811177895437,
            "dram": 1.7200823570792831,
            "package-0": 18.047989094658337,
            "psys": 29.167097595214823,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 683.8374124439363,
          "decode_tokens_per_s": 6.954513862528223,
          "end_to_end_ms": 11979.455717,
          "handoff_estimate_ms": 303.0441995,
          "handoff_plus_first_decode_ms": 426.230754,
          "output_tokens_per_s_e2e": 1.335619946179563,
          "prefill_tokens_per_s": 832.3676017976165,
          "steady_tpot_ms": 123.1865545,
          "steady_tpot_p95_ms": 125.31922225,
          "ttft_ms": 9822.583174
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 967,
          "gross_j": {
            "core": 31.168621,
            "dram": 20.148264,
            "package-0": 216.592219,
            "psys": 336.905937,
            "uncore": 0.0
          },
          "logits": {
            "l2": 766.5680953013031,
            "sha256": "5bf315884e32b1eae4ea75eee38ef4394280af5ea8aefed17fa949ce7ecfd94b",
            "sum": -240251.75810273737,
            "top1": 27
          },
          "net_j": {
            "core": 16.17988377791882,
            "dram": 15.364824483026203,
            "package-0": 184.68431880454725,
            "psys": 258.22639944100393,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 9557622,
          "npu_utilization": 0.9730253061433634,
          "seconds": 9.822583174,
          "watts": {
            "core": 3.1731592848714323,
            "dram": 2.0512184669844977,
            "package-0": 22.05043369582365,
            "psys": 34.29911776077163,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 442.27127637103683,
            "sha256": "d6e71a077cdc3ab7718cb5c7946ef61d3371f158d3d849c96c27ab82187f7a4e",
            "sum": -55858.642204726115,
            "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": 170,
          "gross_j": {
            "core": 7.645915,
            "dram": 5.298449,
            "package-0": 24.7478,
            "psys": 43.804392,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.005048077179126,
            "dram": 4.455654372906234,
            "package-0": 19.125944272307102,
            "psys": 29.941837392244672,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1718228,
          "npu_utilization": 0.9928270604125577,
          "seconds": 1.730641789,
          "token_latencies_s": [
            0.128157366,
            0.123718048,
            0.123790991,
            0.123224991,
            0.123290654,
            0.123148118,
            0.123656615,
            0.123530244,
            0.123034049,
            0.122994115,
            0.12296374,
            0.122956266,
            0.123091253,
            0.123050681
          ],
          "watts": {
            "core": 4.417965085899124,
            "dram": 3.0615515201800085,
            "package-0": 14.299781825041785,
            "psys": 25.311067997098963,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 34.015965,
            "dram": 25.939203,
            "package-0": 237.71039,
            "psys": 387.635421,
            "uncore": 0.0
          },
          "net_j": {
            "core": 15.766608800856353,
            "dram": 20.11518391761104,
            "package-0": 198.86131101449956,
            "psys": 291.8400991377456,
            "uncore": 0.0
          }
        },
        "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": 39,
          "gross_j": {
            "core": 5.235216,
            "dram": 0.71234,
            "package-0": 6.988263,
            "psys": 11.420808,
            "uncore": 0.0
          },
          "logits": {
            "l2": 539.8529262282318,
            "sha256": "21b75f19253b392d43acf6a0728dac2329dc19092f92875416d390c56bfec5b5",
            "sum": -136239.45887422375,
            "top1": 546
          },
          "net_j": {
            "core": 4.632088468458037,
            "dram": 0.5198605385612796,
            "package-0": 5.704330084475936,
            "psys": 8.254844479009567,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 147853,
          "npu_utilization": 0.3740764070764267,
          "seconds": 0.395248129,
          "watts": {
            "core": 13.245390973122102,
            "dram": 1.8022602707880244,
            "package-0": 17.680698496108505,
            "psys": 28.895286687112943,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 684.9860490650676,
          "decode_tokens_per_s": 7.025808287533666,
          "end_to_end_ms": 11959.367656000002,
          "handoff_estimate_ms": 271.4078135,
          "handoff_plus_first_decode_ms": 395.248129,
          "output_tokens_per_s_e2e": 1.3378633770802102,
          "prefill_tokens_per_s": 832.2151978604437,
          "steady_tpot_ms": 123.84031549999999,
          "steady_tpot_p95_ms": 126.18708135000001,
          "ttft_ms": 9824.381988000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 969,
          "gross_j": {
            "core": 22.851504,
            "dram": 19.922556,
            "package-0": 207.619891,
            "psys": 332.988039,
            "uncore": 0.0
          },
          "logits": {
            "l2": 766.5680953013031,
            "sha256": "5bf315884e32b1eae4ea75eee38ef4394280af5ea8aefed17fa949ce7ecfd94b",
            "sum": -240251.75810273737,
            "top1": 27
          },
          "net_j": {
            "core": 7.860021883843624,
            "dram": 15.13824048963232,
            "package-0": 175.70614749667544,
            "psys": 254.2940928225003,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 9529870,
          "npu_utilization": 0.9700223394856051,
          "seconds": 9.824381988,
          "watts": {
            "core": 2.3259991343895203,
            "dram": 2.0278686256636216,
            "package-0": 21.13312483712436,
            "psys": 33.89404436907365,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 442.27127637103683,
            "sha256": "d6e71a077cdc3ab7718cb5c7946ef61d3371f158d3d849c96c27ab82187f7a4e",
            "sum": -55858.642204726115,
            "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": 171,
          "gross_j": {
            "core": 5.929245,
            "dram": 5.304307,
            "package-0": 23.102236,
            "psys": 43.226574,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.2744984485546915,
            "dram": 4.457082889417443,
            "package-0": 17.45083343334818,
            "psys": 29.291161836235727,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1718624,
          "npu_utilization": 0.9878639515865502,
          "seconds": 1.739737539,
          "token_latencies_s": [
            0.128960463,
            0.124082549,
            0.123907004,
            0.123708785,
            0.123736941,
            0.124693722,
            0.124401387,
            0.123773627,
            0.123647812,
            0.124332317,
            0.12411079,
            0.12365935,
            0.123370689,
            0.123311002
          ],
          "watts": {
            "core": 3.4081261495386976,
            "dram": 3.048912195714827,
            "package-0": 13.27915014886622,
            "psys": 24.846606474242435,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 33.828406,
            "dram": 26.150018000000003,
            "package-0": 238.091126,
            "psys": 392.294893,
            "uncore": 0.0
          },
          "net_j": {
            "core": 15.554093200864596,
            "dram": 20.318034378314564,
            "package-0": 199.18891961023888,
            "psys": 296.3685678572622,
            "uncore": 0.0
          }
        },
        "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": 39,
          "gross_j": {
            "core": 5.429125,
            "dram": 0.712644,
            "package-0": 7.181988,
            "psys": 11.765411,
            "uncore": 0.0
          },
          "logits": {
            "l2": 539.8529262282318,
            "sha256": "21b75f19253b392d43acf6a0728dac2329dc19092f92875416d390c56bfec5b5",
            "sum": -136239.45887422375,
            "top1": 546
          },
          "net_j": {
            "core": 4.82348091855534,
            "dram": 0.5193614179199184,
            "package-0": 5.89269787379552,
            "psys": 8.586237494762425,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 147283,
          "npu_utilization": 0.3710859188455758,
          "seconds": 0.396897302,
          "watts": {
            "core": 13.678916366128384,
            "dram": 1.795537526732797,
            "package-0": 18.095330867227712,
            "psys": 29.643464293440825,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 684.050587194936,
          "decode_tokens_per_s": 7.025409807563141,
          "end_to_end_ms": 11975.722488,
          "handoff_estimate_ms": 273.0188295,
          "handoff_plus_first_decode_ms": 396.897302,
          "output_tokens_per_s_e2e": 1.3360363031151095,
          "prefill_tokens_per_s": 830.8423201669978,
          "steady_tpot_ms": 123.8784725,
          "steady_tpot_p95_ms": 125.85470939999999,
          "ttft_ms": 9840.615724
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 971,
          "gross_j": {
            "core": 22.428959,
            "dram": 20.139963,
            "package-0": 207.759478,
            "psys": 337.076224,
            "uncore": 0.0
          },
          "logits": {
            "l2": 766.5680953013031,
            "sha256": "5bf315884e32b1eae4ea75eee38ef4394280af5ea8aefed17fa949ce7ecfd94b",
            "sum": -240251.75810273737,
            "top1": 27
          },
          "net_j": {
            "core": 7.412705069879179,
            "dram": 15.34774192213394,
            "package-0": 175.79300046338588,
            "psys": 258.25224452894514,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 9562541,
          "npu_utilization": 0.9717421417725102,
          "seconds": 9.840615724,
          "watts": {
            "core": 2.2792231328877772,
            "dram": 2.0466161432237637,
            "package-0": 21.112447008097398,
            "psys": 34.253570452701894,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 442.27127637103683,
            "sha256": "d6e71a077cdc3ab7718cb5c7946ef61d3371f158d3d849c96c27ab82187f7a4e",
            "sum": -55858.642204726115,
            "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": 171,
          "gross_j": {
            "core": 5.970322,
            "dram": 5.297411,
            "package-0": 23.14966,
            "psys": 43.453258,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.317907212430078,
            "dram": 4.4509310382607055,
            "package-0": 17.503221273057477,
            "psys": 29.530085833554637,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1718014,
          "npu_utilization": 0.9883814566417313,
          "seconds": 1.738209462,
          "token_latencies_s": [
            0.127366209,
            0.123726367,
            0.123921256,
            0.12388101,
            0.123176337,
            0.123126677,
            0.123637577,
            0.124589797,
            0.123881658,
            0.123826101,
            0.123875935,
            0.125040825,
            0.124345598,
            0.123776723
          ],
          "watts": {
            "core": 3.4347540561253718,
            "dram": 3.0476252234323646,
            "package-0": 13.318107228207001,
            "psys": 24.998861730968994,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 2.6132829897105694,
    "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": 405965,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "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": 0.042768518440425396,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 5.642685,
            "dram": 0.99963,
            "package-0": 9.6724,
            "psys": 15.646383,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.833673686918092,
            "dram": 0.741445696189947,
            "package-0": 7.950183388130849,
            "psys": 11.39968529290373,
            "uncore": 0.0
          }
        },
        "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": 9,
          "gross_j": {
            "core": 2.425347,
            "dram": 0.091979,
            "package-0": 2.718865,
            "psys": 3.689016,
            "uncore": 0.0
          },
          "logits": {
            "l2": 712.0668989995816,
            "sha256": "436eb677d7403000d1f335cdba1e504bf62a2332f96172cb434e632ffe2586d2",
            "sum": 23519.531799525023,
            "top1": 78
          },
          "net_j": {
            "core": 2.2874452373756147,
            "dram": 0.047969639436519926,
            "package-0": 2.425300864916381,
            "psys": 2.9651360125736432,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 19234,
          "npu_utilization": 0.21283307770827353,
          "seconds": 0.090371291,
          "watts": {
            "core": 26.83758274516627,
            "dram": 1.0177900413085832,
            "package-0": 30.085494739695594,
            "psys": 40.82066283638683,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1931.4554380599402,
          "decode_tokens_per_s": 45.951851992589326,
          "end_to_end_ms": 530.1701400000001,
          "handoff_estimate_ms": 73.9033345,
          "handoff_plus_first_decode_ms": 90.371291,
          "output_tokens_per_s_e2e": 30.178991219686566,
          "prefill_tokens_per_s": 4947.445368914313,
          "steady_tpot_ms": 16.4679565,
          "steady_tpot_p95_ms": 18.03229775,
          "ttft_ms": 203.741512
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 0.349059,
            "dram": 0.246459,
            "package-0": 2.067073,
            "psys": 3.967641,
            "uncore": 0.0
          },
          "logits": {
            "l2": 841.2188648681404,
            "sha256": "ce6ed629ef29e138c367d484774de89c27db3dc3473b93905142453d08c5944d",
            "sum": 177284.36455662176,
            "top1": 291
          },
          "net_j": {
            "core": 0.03816033612501596,
            "dram": 0.14724017105401757,
            "package-0": 1.4052344881333418,
            "psys": 2.3356580824959563,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 200620,
          "npu_utilization": 0.9846790574519737,
          "seconds": 0.203741512,
          "watts": {
            "core": 1.7132443780038307,
            "dram": 1.209665117239338,
            "package-0": 10.145566211366882,
            "psys": 19.473895923575945,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 722.9822522062098,
            "sha256": "7a9d0c61379cffa419a00b2eb4ebfdb6f2d1f954050a9188f0d146a26b6a317d",
            "sum": 73133.63508696854,
            "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": 24,
          "gross_j": {
            "core": 2.868279,
            "dram": 0.661192,
            "package-0": 4.886462,
            "psys": 7.989726,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.508068113417462,
            "dram": 0.5462358856994095,
            "package-0": 4.119648035081126,
            "psys": 6.0988911978341305,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 226370,
          "npu_utilization": 0.9589619322020903,
          "seconds": 0.236057337,
          "token_latencies_s": [
            0.018016455,
            0.01806172,
            0.017438221,
            0.017971406,
            0.016983609,
            0.016698959,
            0.016499473,
            0.01643644,
            0.016391744,
            0.016305754,
            0.016296229,
            0.016269944,
            0.016282356,
            0.016375491
          ],
          "watts": {
            "core": 12.15077250490206,
            "dram": 2.8009805092395834,
            "package-0": 20.70031824513889,
            "psys": 33.84654805285717,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 4.278676,
            "dram": 0.958982,
            "package-0": 8.308817,
            "psys": 13.827296,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.58052776258105,
            "dram": 0.7361780494006368,
            "package-0": 6.8226047893824315,
            "psys": 10.162545617504191,
            "uncore": 0.0
          }
        },
        "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": 2,
          "gross_j": {
            "core": 0.663023,
            "dram": 0.049194,
            "package-0": 0.799192,
            "psys": 1.091794,
            "uncore": 0.0
          },
          "logits": {
            "l2": 712.0668989995816,
            "sha256": "436eb677d7403000d1f335cdba1e504bf62a2332f96172cb434e632ffe2586d2",
            "sum": 23519.531799525023,
            "top1": 78
          },
          "net_j": {
            "core": 0.6319698710816558,
            "dram": 0.03928384133934501,
            "package-0": 0.7330863552347182,
            "psys": 0.9287885521857665,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 16716,
          "npu_utilization": 0.8214219937065591,
          "seconds": 0.020350076,
          "watts": {
            "core": 32.58086112307394,
            "dram": 2.4173865493180466,
            "package-0": 39.27218748470521,
            "psys": 53.650610444894646,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2238.162637036616,
          "decode_tokens_per_s": 58.541501709724066,
          "end_to_end_ms": 457.518137,
          "handoff_estimate_ms": 3.706964,
          "handoff_plus_first_decode_ms": 20.350076,
          "output_tokens_per_s_e2e": 34.97129120369713,
          "prefill_tokens_per_s": 5007.708866034781,
          "steady_tpot_ms": 16.643112000000002,
          "steady_tpot_p95_ms": 17.64209375,
          "ttft_ms": 201.289657
        },
        "prefill": {
          "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": 0.641356,
            "dram": 0.24707,
            "package-0": 2.397333,
            "psys": 4.464466,
            "uncore": 0.0
          },
          "logits": {
            "l2": 841.2188648681404,
            "sha256": "ce6ed629ef29e138c367d484774de89c27db3dc3473b93905142453d08c5944d",
            "sum": 177284.36455662176,
            "top1": 291
          },
          "net_j": {
            "core": 0.33419873588215143,
            "dram": 0.1490451848779915,
            "package-0": 1.743459149083305,
            "psys": 2.8521226023803328,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 198320,
          "npu_utilization": 0.9852468475317636,
          "seconds": 0.201289657,
          "watts": {
            "core": 3.186234253456947,
            "dram": 1.2274351483444577,
            "package-0": 11.909866784660476,
            "psys": 22.17931147848297,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 722.9822522062098,
            "sha256": "7a9d0c61379cffa419a00b2eb4ebfdb6f2d1f954050a9188f0d146a26b6a317d",
            "sum": 73133.63508696854,
            "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": 23,
          "gross_j": {
            "core": 2.974297,
            "dram": 0.662718,
            "package-0": 5.112292,
            "psys": 8.271036,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.6143591556172425,
            "dram": 0.5478490231833003,
            "package-0": 4.346059285064408,
            "psys": 6.3816344629380914,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 223247,
          "npu_utilization": 0.9464495104859196,
          "seconds": 0.235878404,
          "token_latencies_s": [
            0.017965628,
            0.017318933,
            0.017467883,
            0.017039755,
            0.017103635,
            0.016740186,
            0.016546038,
            0.016509856,
            0.016455249,
            0.016495377,
            0.016426851,
            0.01649644,
            0.016885393,
            0.016388464
          ],
          "watts": {
            "core": 12.60945024878157,
            "dram": 2.8095747163017095,
            "package-0": 21.673421192047744,
            "psys": 35.06482941948344,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 4.111073,
            "dram": 0.9642919999999999,
            "package-0": 8.127299,
            "psys": 13.704004000000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.407356510257411,
            "dram": 0.7397110219025407,
            "package-0": 6.629233139753828,
            "psys": 10.010024502913383,
            "uncore": 0.0
          }
        },
        "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": 2,
          "gross_j": {
            "core": 0.710509,
            "dram": 0.051452,
            "package-0": 0.812437,
            "psys": 1.164242,
            "uncore": 0.0
          },
          "logits": {
            "l2": 712.0668989995816,
            "sha256": "436eb677d7403000d1f335cdba1e504bf62a2332f96172cb434e632ffe2586d2",
            "sum": 23519.531799525023,
            "top1": 78
          },
          "net_j": {
            "core": 0.6774841741396297,
            "dram": 0.04091260266978655,
            "package-0": 0.7421340230439073,
            "psys": 0.9908866340234442,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 16723,
          "npu_utilization": 0.7727036866416938,
          "seconds": 0.02164219,
          "watts": {
            "core": 32.829810661490356,
            "dram": 2.3773934153613845,
            "package-0": 37.53950039252035,
            "psys": 53.79501797184111,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2220.4528711209146,
          "decode_tokens_per_s": 58.259557701827305,
          "end_to_end_ms": 461.16718499999996,
          "handoff_estimate_ms": 5.040392499999998,
          "handoff_plus_first_decode_ms": 21.64219,
          "output_tokens_per_s_e2e": 34.69457611126429,
          "prefill_tokens_per_s": 4948.485164861213,
          "steady_tpot_ms": 16.6017975,
          "steady_tpot_p95_ms": 17.67764695,
          "ttft_ms": 203.698701
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 21,
          "gross_j": {
            "core": 0.450682,
            "dram": 0.249572,
            "package-0": 2.210688,
            "psys": 4.259754,
            "uncore": 0.0
          },
          "logits": {
            "l2": 841.2188648681404,
            "sha256": "ce6ed629ef29e138c367d484774de89c27db3dc3473b93905142453d08c5944d",
            "sum": 177284.36455662176,
            "top1": 291
          },
          "net_j": {
            "core": 0.1398486634241437,
            "dram": 0.150374019320256,
            "package-0": 1.548988556349011,
            "psys": 2.6281140014229605,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 198779,
          "npu_utilization": 0.9758481474066936,
          "seconds": 0.203698701,
          "watts": {
            "core": 2.2124932451091084,
            "dram": 1.2252017257586731,
            "package-0": 10.852734892992764,
            "psys": 20.91203320928394,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 722.9822522062098,
            "sha256": "7a9d0c61379cffa419a00b2eb4ebfdb6f2d1f954050a9188f0d146a26b6a317d",
            "sum": 73133.63508696854,
            "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": 24,
          "gross_j": {
            "core": 2.949882,
            "dram": 0.663268,
            "package-0": 5.104174,
            "psys": 8.280008,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.5900236726936376,
            "dram": 0.5484243999124981,
            "package-0": 4.33811056036091,
            "psys": 6.391023867466979,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 222688,
          "npu_utilization": 0.9442882565079872,
          "seconds": 0.235826294,
          "token_latencies_s": [
            0.018017478,
            0.017494661,
            0.017361773,
            0.017047193,
            0.01711373,
            0.016748173,
            0.016467834,
            0.016490775,
            0.016367011,
            0.016488662,
            0.016500223,
            0.016572336,
            0.016631259,
            0.016498169
          ],
          "watts": {
            "core": 12.508706938336571,
            "dram": 2.8125277667298625,
            "package-0": 21.64378667630676,
            "psys": 35.11062256696448,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 5.864324741996825,
    "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": 406099,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.038750503212213516,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 6.904523,
            "dram": 1.270443,
            "package-0": 13.941065,
            "psys": 22.660097999999998,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.791324035398854,
            "dram": 0.9151815901406796,
            "package-0": 11.571296224025504,
            "psys": 16.8166451161279,
            "uncore": 0.0
          }
        },
        "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": 2.492852,
            "dram": 0.099609,
            "package-0": 2.859184,
            "psys": 4.017812,
            "uncore": 0.0
          },
          "logits": {
            "l2": 759.4074790825704,
            "sha256": "c51b071629174bb3f22fc600a0fbef27184fb95d6a4b9524fdf7f550639846d2",
            "sum": -148789.21454068646,
            "top1": 522
          },
          "net_j": {
            "core": 2.3351354028972557,
            "dram": 0.049276020517746454,
            "package-0": 2.5234382086796256,
            "psys": 3.189919113390342,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 19207,
          "npu_utilization": 0.1858324142254534,
          "seconds": 0.103356565,
          "watts": {
            "core": 24.11895170858281,
            "dram": 0.9637413936889254,
            "package-0": 27.663303245420355,
            "psys": 38.873312014577884,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2807.349539107639,
          "decode_tokens_per_s": 42.792829269503265,
          "end_to_end_ms": 729.5137179999999,
          "handoff_estimate_ms": 86.4074495,
          "handoff_plus_first_decode_ms": 103.356565,
          "output_tokens_per_s_e2e": 21.93241827427843,
          "prefill_tokens_per_s": 5361.651564518712,
          "steady_tpot_ms": 16.9491155,
          "steady_tpot_p95_ms": 20.958004949999996,
          "ttft_ms": 378.987701
        },
        "prefill": {
          "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": 1.311948,
            "dram": 0.491637,
            "package-0": 5.751328,
            "psys": 9.919469,
            "uncore": 0.0
          },
          "logits": {
            "l2": 659.9256228894352,
            "sha256": "ae1f9856e8f12d8b4cc30ec030d87b4669b23f386612aacb1b94e80973a058b4",
            "sum": -92554.5169948861,
            "top1": 220
          },
          "net_j": {
            "core": 0.7336330129886541,
            "dram": 0.3070761046329815,
            "package-0": 4.520215824658894,
            "psys": 6.883752577241097,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 357540,
          "npu_utilization": 0.9434079234143802,
          "seconds": 0.378987701,
          "watts": {
            "core": 3.4617165584484226,
            "dram": 1.297237347551814,
            "package-0": 15.175500378572972,
            "psys": 26.173590788900032,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 734.1197976827015,
            "sha256": "572b0503dc5eccc03209bbe773e7bd58c98289cecafb7deafd8ec09b9210e019",
            "sum": -60055.14825826511,
            "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": 25,
          "gross_j": {
            "core": 3.099723,
            "dram": 0.679197,
            "package-0": 5.330553,
            "psys": 8.722817,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.7225556195129443,
            "dram": 0.5588294649899516,
            "package-0": 4.527642190686985,
            "psys": 6.74297342549646,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 229064,
          "npu_utilization": 0.9267488281682964,
          "seconds": 0.247169452,
          "token_latencies_s": [
            0.022681491,
            0.020029974,
            0.017266856,
            0.017665278,
            0.017142405,
            0.016938804,
            0.016902463,
            0.016959192,
            0.016924811,
            0.016939039,
            0.016970031,
            0.016846037,
            0.016928737,
            0.016935161
          ],
          "watts": {
            "core": 12.540882276989471,
            "dram": 2.747900254275759,
            "package-0": 21.56639081758372,
            "psys": 35.29083763959633,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 5.409472,
            "dram": 1.235836,
            "package-0": 12.384001,
            "psys": 19.911143,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.405662745547006,
            "dram": 0.9154847399765256,
            "package-0": 10.247100136919109,
            "psys": 14.641903385774786,
            "uncore": 0.0
          }
        },
        "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": 4,
          "gross_j": {
            "core": 1.071653,
            "dram": 0.063049,
            "package-0": 1.281186,
            "psys": 1.72845,
            "uncore": 0.0
          },
          "logits": {
            "l2": 759.4074790825704,
            "sha256": "c51b071629174bb3f22fc600a0fbef27184fb95d6a4b9524fdf7f550639846d2",
            "sum": -148789.21454068646,
            "top1": 522
          },
          "net_j": {
            "core": 1.0231790880922187,
            "dram": 0.0475792494582134,
            "package-0": 1.1779951360653067,
            "psys": 1.4739986131973049,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 16491,
          "npu_utilization": 0.5191325415169096,
          "seconds": 0.031766454,
          "watts": {
            "core": 33.735367504349085,
            "dram": 1.984766697598668,
            "package-0": 40.33141376119601,
            "psys": 54.41117223848781,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3113.2793270680786,
          "decode_tokens_per_s": 54.76064976081041,
          "end_to_end_ms": 657.827257,
          "handoff_estimate_ms": 14.449682999999999,
          "handoff_plus_first_decode_ms": 31.766454,
          "output_tokens_per_s_e2e": 24.322494742719364,
          "prefill_tokens_per_s": 5292.935647988272,
          "steady_tpot_ms": 17.316771000000003,
          "steady_tpot_p95_ms": 17.806366,
          "ttft_ms": 383.907936
        },
        "prefill": {
          "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": 1.25726,
            "dram": 0.494872,
            "package-0": 5.71727,
            "psys": 9.543555,
            "uncore": 0.0
          },
          "logits": {
            "l2": 659.9256228894352,
            "sha256": "ae1f9856e8f12d8b4cc30ec030d87b4669b23f386612aacb1b94e80973a058b4",
            "sum": -92554.5169948861,
            "top1": 220
          },
          "net_j": {
            "core": 0.6714369972071138,
            "dram": 0.30791502957666683,
            "package-0": 4.470174822577147,
            "psys": 6.468427171134984,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 357996,
          "npu_utilization": 0.9325048180301226,
          "seconds": 0.383907936,
          "watts": {
            "core": 3.2748997405461293,
            "dram": 1.289038213578372,
            "package-0": 14.892294385912356,
            "psys": 24.858967750018067,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 734.1197976827015,
            "sha256": "572b0503dc5eccc03209bbe773e7bd58c98289cecafb7deafd8ec09b9210e019",
            "sum": -60055.14825826511,
            "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": 24,
          "gross_j": {
            "core": 3.080559,
            "dram": 0.677915,
            "package-0": 5.385545,
            "psys": 8.639138,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.7110466602476735,
            "dram": 0.5599904609416454,
            "package-0": 4.598930178276654,
            "psys": 6.699477601442497,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 227520,
          "npu_utilization": 0.9395717788466242,
          "seconds": 0.242152867,
          "token_latencies_s": [
            0.017941085,
            0.017693686,
            0.017619532,
            0.017454019,
            0.017733825,
            0.017389973,
            0.017064191,
            0.016913076,
            0.016920434,
            0.017038488,
            0.016919895,
            0.016798849,
            0.017255338,
            0.017378204
          ],
          "watts": {
            "core": 12.721546674894418,
            "dram": 2.7995332386463136,
            "package-0": 22.24026940800168,
            "psys": 35.67638123400786,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 5.414354,
            "dram": 1.22662,
            "package-0": 12.274626,
            "psys": 19.64979,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.427842732563086,
            "dram": 0.9117891432920843,
            "package-0": 10.17454894903306,
            "psys": 14.471351738748112,
            "uncore": 0.0
          }
        },
        "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": 3,
          "gross_j": {
            "core": 1.007322,
            "dram": 0.05957,
            "package-0": 1.20477,
            "psys": 1.613765,
            "uncore": 0.0
          },
          "logits": {
            "l2": 759.4074790825704,
            "sha256": "c51b071629174bb3f22fc600a0fbef27184fb95d6a4b9524fdf7f550639846d2",
            "sum": -148789.21454068646,
            "top1": 522
          },
          "net_j": {
            "core": 0.960338896702405,
            "dram": 0.044576019545616324,
            "package-0": 1.1047527571408875,
            "psys": 1.3671392312130513,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 16489,
          "npu_utilization": 0.5355400436921948,
          "seconds": 0.030789481,
          "watts": {
            "core": 32.71643325199278,
            "dram": 1.934751677041909,
            "package-0": 39.12927275389929,
            "psys": 52.41286788822455,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3167.8691398301476,
          "decode_tokens_per_s": 55.144178959330446,
          "end_to_end_ms": 646.491351,
          "handoff_estimate_ms": 13.804062499999999,
          "handoff_plus_first_decode_ms": 30.789481000000002,
          "output_tokens_per_s_e2e": 24.748977654923028,
          "prefill_tokens_per_s": 5426.232339422924,
          "steady_tpot_ms": 16.9854185,
          "steady_tpot_p95_ms": 17.955736849999997,
          "ttft_ms": 374.477146
        },
        "prefill": {
          "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": 1.403622,
            "dram": 0.488829,
            "package-0": 5.802841,
            "psys": 9.567053,
            "uncore": 0.0
          },
          "logits": {
            "l2": 659.9256228894352,
            "sha256": "ae1f9856e8f12d8b4cc30ec030d87b4669b23f386612aacb1b94e80973a058b4",
            "sum": -92554.5169948861,
            "top1": 220
          },
          "net_j": {
            "core": 0.8321898789744265,
            "dram": 0.30646467213954076,
            "package-0": 4.5863810130106355,
            "psys": 6.567466419589879,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 359521,
          "npu_utilization": 0.9600612583177506,
          "seconds": 0.374477146,
          "watts": {
            "core": 3.7482180554751396,
            "dram": 1.3053640394920123,
            "package-0": 15.495848176539992,
            "psys": 25.54776199880566,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 734.1197976827015,
            "sha256": "572b0503dc5eccc03209bbe773e7bd58c98289cecafb7deafd8ec09b9210e019",
            "sum": -60055.14825826511,
            "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": 24,
          "gross_j": {
            "core": 3.00341,
            "dram": 0.678221,
            "package-0": 5.267015,
            "psys": 8.468972,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.6353139568862542,
            "dram": 0.5607484516069272,
            "package-0": 4.483415178881536,
            "psys": 6.536746087945183,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 228114,
          "npu_utilization": 0.9456493356792068,
          "seconds": 0.241224724,
          "token_latencies_s": [
            0.017436233,
            0.017452104,
            0.017352495,
            0.01732255,
            0.018891055,
            0.016965704,
            0.017005133,
            0.016849093,
            0.016948205,
            0.016854944,
            0.016932022,
            0.017425609,
            0.016885595,
            0.016874093
          ],
          "watts": {
            "core": 12.450672344846376,
            "dram": 2.8115733277821056,
            "package-0": 21.834474147848958,
            "psys": 35.10822547360444,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 14.634349533356726,
    "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": 406304,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.03937573544681072,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 7.848979,
            "dram": 2.249933,
            "package-0": 23.131471,
            "psys": 40.520343,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.858397596178624,
            "dram": 1.614667626559138,
            "package-0": 18.89393772264779,
            "psys": 30.0712956916333,
            "uncore": 0.0
          }
        },
        "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": 12,
          "gross_j": {
            "core": 2.589349,
            "dram": 0.123657,
            "package-0": 3.01934,
            "psys": 4.441944,
            "uncore": 0.0
          },
          "logits": {
            "l2": 745.9631922969153,
            "sha256": "24b1cbf9fe48ed8a6823bcda5c695ed01600d1970212ae8682be6ad913cc2694",
            "sum": -47925.70799851045,
            "top1": 78
          },
          "net_j": {
            "core": 2.40352184985598,
            "dram": 0.06435294254971488,
            "package-0": 2.6237526951758303,
            "psys": 3.466491964074085,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 22063,
          "npu_utilization": 0.18117352289751504,
          "seconds": 0.12177828,
          "watts": {
            "core": 21.262814682552584,
            "dram": 1.0154273816316013,
            "package-0": 24.79374811337457,
            "psys": 36.47566708940215,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3139.9254451073525,
          "decode_tokens_per_s": 36.15686424147231,
          "end_to_end_ms": 1304.489572,
          "handoff_estimate_ms": 102.36964700000001,
          "handoff_plus_first_decode_ms": 121.77828000000001,
          "output_tokens_per_s_e2e": 12.265333769950596,
          "prefill_tokens_per_s": 4586.173249459104,
          "steady_tpot_ms": 19.408633000000002,
          "steady_tpot_p95_ms": 27.549739649999996,
          "ttft_ms": 889.630587
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 88,
          "gross_j": {
            "core": 2.150263,
            "dram": 1.352658,
            "package-0": 14.464135,
            "psys": 26.18548,
            "uncore": 0.0
          },
          "logits": {
            "l2": 667.2018758805161,
            "sha256": "d7b52e24c142f2b72de7a972f3de94ad4cd10f7c3d5413ce59a0998a819faea1",
            "sum": 10659.559723772109,
            "top1": 220
          },
          "net_j": {
            "core": 0.792734246406487,
            "dram": 0.9194222584459324,
            "package-0": 11.574238984056318,
            "psys": 19.059480459596593,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 853218,
          "npu_utilization": 0.9590699920482838,
          "seconds": 0.889630587,
          "watts": {
            "core": 2.417029080858255,
            "dram": 1.5204715527614834,
            "package-0": 16.25858554254048,
            "psys": 29.434104877511366,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 722.5170187115344,
            "sha256": "d8ec2f43616339a09570ca4449bf8889855b4f8bd315d2af2e02def48909cafd",
            "sum": 16701.138758618385,
            "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": 29,
          "gross_j": {
            "core": 3.109367,
            "dram": 0.773618,
            "package-0": 5.647996,
            "psys": 9.892919,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.662141499916157,
            "dram": 0.6308924255634908,
            "package-0": 4.695946043415644,
            "psys": 7.5453232679626225,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 266669,
          "npu_utilization": 0.9098824844167069,
          "seconds": 0.293080705,
          "token_latencies_s": [
            0.035609648,
            0.023209789,
            0.019711727,
            0.019431841,
            0.019300816,
            0.019221497,
            0.019385425,
            0.021546063,
            0.019436998,
            0.019447245,
            0.019148993,
            0.019087076,
            0.019229221,
            0.019285769
          ],
          "watts": {
            "core": 10.609251810009125,
            "dram": 2.639607407795747,
            "package-0": 19.271128749332032,
            "psys": 33.75493108630266,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.760175,
            "dram": 2.191644,
            "package-0": 21.855168,
            "psys": 37.534449,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.903234967827878,
            "dram": 1.5990283447293832,
            "package-0": 17.90212937255638,
            "psys": 27.786917847685658,
            "uncore": 0.0
          }
        },
        "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": 5,
          "gross_j": {
            "core": 1.721004,
            "dram": 0.08258,
            "package-0": 1.989985,
            "psys": 2.685662,
            "uncore": 0.0
          },
          "logits": {
            "l2": 745.9631922969153,
            "sha256": "24b1cbf9fe48ed8a6823bcda5c695ed01600d1970212ae8682be6ad913cc2694",
            "sum": -47925.70799851045,
            "top1": 78
          },
          "net_j": {
            "core": 1.6423635326121473,
            "dram": 0.05748302793605765,
            "package-0": 1.8225758215243928,
            "psys": 2.2728590056033546,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 19879,
          "npu_utilization": 0.3857338735871374,
          "seconds": 0.051535531,
          "watts": {
            "core": 33.39451377730055,
            "dram": 1.6023896212498518,
            "package-0": 38.613844883057475,
            "psys": 52.112822898826835,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3365.9014788458758,
          "decode_tokens_per_s": 45.792923101217276,
          "end_to_end_ms": 1216.910247,
          "handoff_estimate_ms": 31.905805500000003,
          "handoff_plus_first_decode_ms": 51.535531000000006,
          "output_tokens_per_s_e2e": 13.148052651741702,
          "prefill_tokens_per_s": 4587.626795426779,
          "steady_tpot_ms": 19.6297255,
          "steady_tpot_p95_ms": 20.2453457,
          "ttft_ms": 889.348716
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 89,
          "gross_j": {
            "core": 1.860774,
            "dram": 1.351315,
            "package-0": 14.14187,
            "psys": 25.090573,
            "uncore": 0.0
          },
          "logits": {
            "l2": 667.2018758805161,
            "sha256": "d7b52e24c142f2b72de7a972f3de94ad4cd10f7c3d5413ce59a0998a819faea1",
            "sum": 10659.559723772109,
            "top1": 220
          },
          "net_j": {
            "core": 0.5036753664955396,
            "dram": 0.9182165250749041,
            "package-0": 11.252889620154958,
            "psys": 17.966831264853973,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 853201,
          "npu_utilization": 0.9593548454619909,
          "seconds": 0.889348716,
          "watts": {
            "core": 2.0922883977042814,
            "dram": 1.519443358593661,
            "package-0": 15.90137787976522,
            "psys": 28.21230024691462,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 722.5170187115344,
            "sha256": "d8ec2f43616339a09570ca4449bf8889855b4f8bd315d2af2e02def48909cafd",
            "sum": 16701.138758618385,
            "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": 28,
          "gross_j": {
            "core": 3.178397,
            "dram": 0.757749,
            "package-0": 5.723313,
            "psys": 9.758214,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.7571960687201913,
            "dram": 0.6233287917184214,
            "package-0": 4.826663930877031,
            "psys": 7.547227577228332,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 260737,
          "npu_utilization": 0.9446102903349685,
          "seconds": 0.276026,
          "token_latencies_s": [
            0.020099603,
            0.020225822,
            0.019960102,
            0.020281604,
            0.019838353,
            0.019528373,
            0.019682196,
            0.019430199,
            0.019438017,
            0.019428453,
            0.019577255,
            0.01969393,
            0.019471025,
            0.019318372
          ],
          "watts": {
            "core": 11.514846427510452,
            "dram": 2.745208784679704,
            "package-0": 20.734688036634232,
            "psys": 35.352517516465845,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.955915,
            "dram": 2.1974430000000003,
            "package-0": 22.093084,
            "psys": 37.26321,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.084102852864798,
            "dram": 1.6000811235260004,
            "package-0": 18.10838573686418,
            "psys": 27.437611488809317,
            "uncore": 0.0
          }
        },
        "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": 1.802425,
            "dram": 0.085083,
            "package-0": 2.040461,
            "psys": 2.784172,
            "uncore": 0.0
          },
          "logits": {
            "l2": 745.9631922969153,
            "sha256": "24b1cbf9fe48ed8a6823bcda5c695ed01600d1970212ae8682be6ad913cc2694",
            "sum": -47925.70799851045,
            "top1": 78
          },
          "net_j": {
            "core": 1.7190717946100482,
            "dram": 0.05848202550687543,
            "package-0": 1.8630193837285818,
            "psys": 2.346630694353518,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 18781,
          "npu_utilization": 0.3438236433135702,
          "seconds": 0.054623934,
          "watts": {
            "core": 32.99698260473147,
            "dram": 1.5576139206670836,
            "package-0": 37.354706089092744,
            "psys": 50.969818468219444,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 3339.1583711978074,
          "decode_tokens_per_s": 45.60808645967518,
          "end_to_end_ms": 1226.656404,
          "handoff_estimate_ms": 35.06952,
          "handoff_plus_first_decode_ms": 54.623934,
          "output_tokens_per_s_e2e": 13.043587387491435,
          "prefill_tokens_per_s": 4544.607195949524,
          "steady_tpot_ms": 19.554413999999998,
          "steady_tpot_p95_ms": 20.106267550000002,
          "ttft_ms": 897.767359
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 89,
          "gross_j": {
            "core": 2.179926,
            "dram": 1.359066,
            "package-0": 14.515527,
            "psys": 25.170407,
            "uncore": 0.0
          },
          "logits": {
            "l2": 667.2018758805161,
            "sha256": "d7b52e24c142f2b72de7a972f3de94ad4cd10f7c3d5413ce59a0998a819faea1",
            "sum": 10659.559723772109,
            "top1": 220
          },
          "net_j": {
            "core": 0.8099809669834928,
            "dram": 0.9218677817698958,
            "package-0": 11.599199310810082,
            "psys": 17.979231380469308,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 852678,
          "npu_utilization": 0.9497761212323158,
          "seconds": 0.897767359,
          "watts": {
            "core": 2.4281635750582016,
            "dram": 1.5138287067084157,
            "package-0": 16.168472661078336,
            "psys": 28.03666979832801,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 722.5170187115344,
            "sha256": "d8ec2f43616339a09570ca4449bf8889855b4f8bd315d2af2e02def48909cafd",
            "sum": 16701.138758618385,
            "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": 2.973564,
            "dram": 0.753294,
            "package-0": 5.537096,
            "psys": 9.308631,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.555050091271257,
            "dram": 0.6197313162492292,
            "package-0": 4.646167042325514,
            "psys": 7.111749413986491,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 260537,
          "npu_utilization": 0.9499458354365752,
          "seconds": 0.274265111,
          "token_latencies_s": [
            0.019712694,
            0.020071207,
            0.019838815,
            0.02017138,
            0.019883847,
            0.019660559,
            0.019476924,
            0.019548802,
            0.019560026,
            0.019497376,
            0.019239628,
            0.019085801,
            0.019191579,
            0.019298435
          ],
          "watts": {
            "core": 10.84193315423193,
            "dram": 2.7465906883066875,
            "package-0": 20.18884567494259,
            "psys": 33.940266649519266,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 50.01766771264374,
    "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": 406493,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.038754516281187534,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.620087999999999,
            "dram": 5.957077,
            "package-0": 53.325058999999996,
            "psys": 96.75225400000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.3247959297586265,
            "dram": 4.267160833787196,
            "package-0": 42.05248490789287,
            "psys": 68.9559743565998,
            "uncore": 0.0
          }
        },
        "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": 15,
          "gross_j": {
            "core": 2.706902,
            "dram": 0.157348,
            "package-0": 3.196037,
            "psys": 4.814135,
            "uncore": 0.0
          },
          "logits": {
            "l2": 772.4011392999273,
            "sha256": "c1b35fe6f6b041dcd0c1e661cd3228b8443fe53355bb70c15d60c562f9bd4d03",
            "sum": 11189.117347188294,
            "top1": 78
          },
          "net_j": {
            "core": 2.483511991617699,
            "dram": 0.08605629757300691,
            "package-0": 2.7204861923160086,
            "psys": 3.6415063592812436,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 25533,
          "npu_utilization": 0.1744124294714231,
          "seconds": 0.146394383,
          "watts": {
            "core": 18.49047719269393,
            "dram": 1.0748226590087135,
            "package-0": 21.831691452260163,
            "psys": 32.88469749553165,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2360.692145969335,
          "decode_tokens_per_s": 30.53897379764192,
          "end_to_end_ms": 3470.1687019999995,
          "handoff_estimate_ms": 123.1576245,
          "handoff_plus_first_decode_ms": 146.39438299999998,
          "output_tokens_per_s_e2e": 4.610726847596357,
          "prefill_tokens_per_s": 2744.551541355688,
          "steady_tpot_ms": 23.2367585,
          "steady_tpot_p95_ms": 30.86077999999999,
          "ttft_ms": 2978.993062
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 294,
          "gross_j": {
            "core": 5.745225,
            "dram": 4.937304,
            "package-0": 44.037546,
            "psys": 81.205786,
            "uncore": 0.0
          },
          "logits": {
            "l2": 674.7263359808221,
            "sha256": "dcb96fa05f02a6152ed6574236449268a41f7b136d2e8fd7a99247570c66fbf0",
            "sum": -598.7580670975149,
            "top1": 198
          },
          "net_j": {
            "core": 1.1994407188435536,
            "dram": 3.486582445279004,
            "package-0": 34.36051790856628,
            "psys": 57.34385555282111,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2901686,
          "npu_utilization": 0.9740492641671015,
          "seconds": 2.978993062,
          "watts": {
            "core": 1.9285795167790156,
            "dram": 1.6573734470818988,
            "package-0": 14.782695052815804,
            "psys": 27.259474698299922,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 723.3369122262443,
            "sha256": "1883b45b8d5aa18bcfafb26d9ce86cb59c90e61fe4ce23a8ce37ca7672e44044",
            "sum": 66699.60143572465,
            "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": 33,
          "gross_j": {
            "core": 3.167961,
            "dram": 0.862425,
            "package-0": 6.091476,
            "psys": 10.732333,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.6418432192973738,
            "dram": 0.6945220909351855,
            "package-0": 4.971480807010582,
            "psys": 7.970612444497441,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 307694,
          "npu_utilization": 0.8924325025011438,
          "seconds": 0.344781257,
          "token_latencies_s": [
            0.042751295,
            0.024458195,
            0.023101537,
            0.023272606,
            0.023248144,
            0.023309277,
            0.022952439,
            0.022673416,
            0.023052258,
            0.02340053,
            0.02320157,
            0.022706691,
            0.023225373,
            0.023402395
          ],
          "watts": {
            "core": 9.188321394164417,
            "dram": 2.5013685706238955,
            "package-0": 17.667654132370657,
            "psys": 31.127947886099854,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.757297,
            "dram": 5.951827,
            "package-0": 52.352711,
            "psys": 91.399119,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.5860765730954345,
            "dram": 4.30150651105338,
            "package-0": 41.344259599085596,
            "psys": 64.2541216746021,
            "uncore": 0.0
          }
        },
        "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": 8,
          "gross_j": {
            "core": 2.327997,
            "dram": 0.106506,
            "package-0": 2.663995,
            "psys": 3.500235,
            "uncore": 0.0
          },
          "logits": {
            "l2": 772.4011392999273,
            "sha256": "c1b35fe6f6b041dcd0c1e661cd3228b8443fe53355bb70c15d60c562f9bd4d03",
            "sum": 11189.117347188294,
            "top1": 78
          },
          "net_j": {
            "core": 2.2118441664203226,
            "dram": 0.06943749676845401,
            "package-0": 2.4167298057166904,
            "psys": 2.8905204458778,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 22482,
          "npu_utilization": 0.2953550938210918,
          "seconds": 0.076118545,
          "watts": {
            "core": 30.583834727792024,
            "dram": 1.3992122419050443,
            "package-0": 34.99797585463569,
            "psys": 45.983997723550814,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2417.3315714401288,
          "decode_tokens_per_s": 37.76794924809452,
          "end_to_end_ms": 3388.860716,
          "handoff_estimate_ms": 53.518238499999995,
          "handoff_plus_first_decode_ms": 76.118545,
          "output_tokens_per_s_e2e": 4.7213507254690015,
          "prefill_tokens_per_s": 2732.8956527854257,
          "steady_tpot_ms": 22.600306500000002,
          "steady_tpot_p95_ms": 24.377819849999995,
          "ttft_ms": 2991.698564
        },
        "prefill": {
          "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": 5.476915,
            "dram": 4.999193,
            "package-0": 43.832713,
            "psys": 78.008528,
            "uncore": 0.0
          },
          "logits": {
            "l2": 674.7263359808221,
            "sha256": "dcb96fa05f02a6152ed6574236449268a41f7b136d2e8fd7a99247570c66fbf0",
            "sum": -598.7580670975149,
            "top1": 198
          },
          "net_j": {
            "core": 0.9117428014923714,
            "dram": 3.54228407081007,
            "package-0": 34.114412069989335,
            "psys": 54.04482564576706,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2903866,
          "npu_utilization": 0.9706412387073633,
          "seconds": 2.991698564,
          "watts": {
            "core": 1.8307041578003043,
            "dram": 1.6710216263619533,
            "package-0": 14.65144701657182,
            "psys": 26.074995970081964,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 723.3369122262443,
            "sha256": "1883b45b8d5aa18bcfafb26d9ce86cb59c90e61fe4ce23a8ce37ca7672e44044",
            "sum": 66699.60143572465,
            "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": 32,
          "gross_j": {
            "core": 2.952385,
            "dram": 0.846128,
            "package-0": 5.856003,
            "psys": 9.890356,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.46248960518274,
            "dram": 0.6897849434748564,
            "package-0": 4.813117723379572,
            "psys": 7.318775582957246,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 305229,
          "npu_utilization": 0.950740003366583,
          "seconds": 0.321043607,
          "token_latencies_s": [
            0.025640621,
            0.02369785,
            0.023048045,
            0.02292465,
            0.022585962,
            0.022614651,
            0.022536019,
            0.022500048,
            0.022584317,
            0.022388062,
            0.022579137,
            0.022416491,
            0.022754714,
            0.022728235
          ],
          "watts": {
            "core": 9.196211778171307,
            "dram": 2.635554739453198,
            "package-0": 18.24052207337678,
            "psys": 30.806892846802587,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.391271,
            "dram": 5.9081269999999995,
            "package-0": 51.835072000000004,
            "psys": 88.98774800000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.232744156687014,
            "dram": 4.261857485359242,
            "package-0": 40.853642595159165,
            "psys": 61.90938239094715,
            "uncore": 0.0
          }
        },
        "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": 8,
          "gross_j": {
            "core": 2.391901,
            "dram": 0.117187,
            "package-0": 2.753411,
            "psys": 3.73583,
            "uncore": 0.0
          },
          "logits": {
            "l2": 772.4011392999273,
            "sha256": "c1b35fe6f6b041dcd0c1e661cd3228b8443fe53355bb70c15d60c562f9bd4d03",
            "sum": 11189.117347188294,
            "top1": 78
          },
          "net_j": {
            "core": 2.25839679721496,
            "dram": 0.07458105729635098,
            "package-0": 2.469208354038404,
            "psys": 3.0350338767725575,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 22993,
          "npu_utilization": 0.26280887815550386,
          "seconds": 0.087489434,
          "watts": {
            "core": 27.339312767756613,
            "dram": 1.3394417433309718,
            "package-0": 31.471354586657856,
            "psys": 42.70035625101883,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 2423.27989763919,
          "decode_tokens_per_s": 36.81603411132208,
          "end_to_end_ms": 3380.542218,
          "handoff_estimate_ms": 64.87825450000001,
          "handoff_plus_first_decode_ms": 87.489434,
          "output_tokens_per_s_e2e": 4.732968550076543,
          "prefill_tokens_per_s": 2749.981376074025,
          "steady_tpot_ms": 22.611179500000002,
          "steady_tpot_p95_ms": 23.9025259,
          "ttft_ms": 2973.1110439999998
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 294,
          "gross_j": {
            "core": 5.07561,
            "dram": 4.948535,
            "package-0": 43.280346,
            "psys": 75.518178,
            "uncore": 0.0
          },
          "logits": {
            "l2": 674.7263359808221,
            "sha256": "dcb96fa05f02a6152ed6574236449268a41f7b136d2e8fd7a99247570c66fbf0",
            "sum": -598.7580670975149,
            "top1": 198
          },
          "net_j": {
            "core": 0.538801364106698,
            "dram": 3.500677893137814,
            "package-0": 33.62242518818583,
            "psys": 51.70336290404346,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2893619,
          "npu_utilization": 0.9732630087394746,
          "seconds": 2.973111044,
          "watts": {
            "core": 1.7071713517875589,
            "dram": 1.6644299276969756,
            "package-0": 14.557258494378658,
            "psys": 25.40038931690841,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 723.3369122262443,
            "sha256": "1883b45b8d5aa18bcfafb26d9ce86cb59c90e61fe4ce23a8ce37ca7672e44044",
            "sum": 66699.60143572465,
            "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": 32,
          "gross_j": {
            "core": 2.92376,
            "dram": 0.842405,
            "package-0": 5.801315,
            "psys": 9.73374,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.4355459953653553,
            "dram": 0.686598534925077,
            "package-0": 4.76200905293493,
            "psys": 7.170985610131134,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 306017,
          "npu_utilization": 0.9564772636418118,
          "seconds": 0.31994174,
          "token_latencies_s": [
            0.025090631,
            0.022843426,
            0.023262777,
            0.022992873,
            0.022606033,
            0.022616326,
            0.02258021,
            0.02253286,
            0.022556541,
            0.022465176,
            0.022628597,
            0.022443708,
            0.022702033,
            0.022571763
          ],
          "watts": {
            "core": 9.138413762455627,
            "dram": 2.632994994651214,
            "package-0": 18.13241060700614,
            "psys": 30.42347647418558,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 4.496121177449822,
    "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": 406788,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "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": 0.0495118023827672,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 7.23863,
            "dram": 2.913627,
            "package-0": 21.426334999999998,
            "psys": 35.955475,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.303336961309061,
            "dram": 2.2960061116598705,
            "package-0": 17.306499138408263,
            "psys": 25.796649805504725,
            "uncore": 0.0
          }
        },
        "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": 12,
          "gross_j": {
            "core": 2.576898,
            "dram": 0.203368,
            "package-0": 3.225699,
            "psys": 4.606434,
            "uncore": 0.0
          },
          "logits": {
            "l2": 992.2683369200568,
            "sha256": "ae79c6c31d292f024b23dc06077f4b134ca9d7fc1fefa9233a47dedcd490f0e2",
            "sum": 292038.3226532545,
            "top1": 78
          },
          "net_j": {
            "core": 2.384725661830048,
            "dram": 0.1420389672168258,
            "package-0": 2.816604111231097,
            "psys": 3.597674524480727,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 50545,
          "npu_utilization": 0.40135313296597036,
          "seconds": 0.125936478,
          "watts": {
            "core": 20.461887142818146,
            "dram": 1.6148458590369663,
            "package-0": 25.613698677518997,
            "psys": 36.57744025523725,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 807.407079374919,
          "decode_tokens_per_s": 18.812497916437472,
          "end_to_end_ms": 1268.257396,
          "handoff_estimate_ms": 78.194027,
          "handoff_plus_first_decode_ms": 125.936478,
          "output_tokens_per_s_e2e": 12.61573561523311,
          "prefill_tokens_per_s": 2140.5131580879497,
          "steady_tpot_ms": 47.742450999999996,
          "steady_tpot_p95_ms": 49.104597749999996,
          "ttft_ms": 470.915115
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 47,
          "gross_j": {
            "core": 0.816221,
            "dram": 0.683531,
            "package-0": 7.808085,
            "psys": 12.827421,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1147.1172002671653,
            "sha256": "d036e9984fc486f880661b10e94e4628553614043ee547042d906b12c17261b9",
            "sum": 371321.3503918741,
            "top1": 291
          },
          "net_j": {
            "core": 0.09762968979104047,
            "dram": 0.4542029372767654,
            "package-0": 6.2783537446085305,
            "psys": 9.055359942190375,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 464276,
          "npu_utilization": 0.9859016735956756,
          "seconds": 0.470915115,
          "watts": {
            "core": 1.7332656650870082,
            "dram": 1.451495138354181,
            "package-0": 16.580663374969394,
            "psys": 27.239348645668336,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 953.3553678944744,
            "sha256": "d7f745389327c7b4f7c74a1b3a7960947e9141430a397b0cdfdaccad2789a89f",
            "sum": 273382.66879007034,
            "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": 66,
          "gross_j": {
            "core": 3.845511,
            "dram": 2.026728,
            "package-0": 10.392551,
            "psys": 18.52162,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.8209816096879723,
            "dram": 1.6997642071662793,
            "package-0": 8.211541282568636,
            "psys": 13.143615338833623,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 654689,
          "npu_utilization": 0.9751017895208749,
          "seconds": 0.671405803,
          "token_latencies_s": [
            0.050040757,
            0.048089667,
            0.047947045,
            0.048600512,
            0.04754948,
            0.048062948,
            0.047716527,
            0.047455843,
            0.047546194,
            0.047806947,
            0.047498061,
            0.04775913,
            0.047725772,
            0.04757785
          ],
          "watts": {
            "core": 5.7275510322033965,
            "dram": 3.0186334269738206,
            "package-0": 15.478792339243453,
            "psys": 27.586326953447557,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 6.3616779999999995,
            "dram": 2.8526540000000002,
            "package-0": 20.359262,
            "psys": 34.282933,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.552548215048158,
            "dram": 2.275296296248665,
            "package-0": 16.508001432577156,
            "psys": 24.786369496670208,
            "uncore": 0.0
          }
        },
        "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": 5,
          "gross_j": {
            "core": 1.645992,
            "dram": 0.149291,
            "package-0": 2.125544,
            "psys": 2.888604,
            "uncore": 0.0
          },
          "logits": {
            "l2": 992.2683369200568,
            "sha256": "ae79c6c31d292f024b23dc06077f4b134ca9d7fc1fefa9233a47dedcd490f0e2",
            "sum": 292038.3226532545,
            "top1": 78
          },
          "net_j": {
            "core": 1.5653396651263465,
            "dram": 0.12355196941625678,
            "package-0": 1.9538519746112566,
            "psys": 2.465240222458822,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 46679,
          "npu_utilization": 0.8831692302394692,
          "seconds": 0.05285397,
          "watts": {
            "core": 31.142258566385834,
            "dram": 2.8245938762972775,
            "package-0": 40.215408606013895,
            "psys": 54.65254549469037,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 863.7132134473471,
          "decode_tokens_per_s": 20.989015272922824,
          "end_to_end_ms": 1185.5787129999999,
          "handoff_estimate_ms": 5.7113035000000005,
          "handoff_plus_first_decode_ms": 52.85397,
          "output_tokens_per_s_e2e": 13.495518960114799,
          "prefill_tokens_per_s": 2140.494712879316,
          "steady_tpot_ms": 47.1426665,
          "steady_tpot_p95_ms": 47.920721449999995,
          "ttft_ms": 470.919173
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 47,
          "gross_j": {
            "core": 0.88751,
            "dram": 0.684691,
            "package-0": 7.899272,
            "psys": 12.93057,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1147.1172002671653,
            "sha256": "d036e9984fc486f880661b10e94e4628553614043ee547042d906b12c17261b9",
            "sum": 371321.3503918741,
            "top1": 291
          },
          "net_j": {
            "core": 0.1689124974998102,
            "dram": 0.45536096109624186,
            "package-0": 6.369527562510022,
            "psys": 9.158476437343356,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 466181,
          "npu_utilization": 0.9899384580801512,
          "seconds": 0.470919173,
          "watts": {
            "core": 1.8846333954638113,
            "dram": 1.4539458982698927,
            "package-0": 16.774156698011527,
            "psys": 27.458151507456247,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 953.3553678944744,
            "sha256": "d7f745389327c7b4f7c74a1b3a7960947e9141430a397b0cdfdaccad2789a89f",
            "sum": 273382.66879007034,
            "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": 66,
          "gross_j": {
            "core": 3.828176,
            "dram": 2.018672,
            "package-0": 10.334446,
            "psys": 18.463759,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.818296052422001,
            "dram": 1.6963833657361667,
            "package-0": 8.184621895455876,
            "psys": 13.162652836868029,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 651910,
          "npu_utilization": 0.9850476175351622,
          "seconds": 0.66180557,
          "token_latencies_s": [
            0.048738439,
            0.047480412,
            0.047287563,
            0.047167345,
            0.047151295,
            0.047140702,
            0.047061307,
            0.047381115,
            0.047144631,
            0.047137891,
            0.047034206,
            0.046999555,
            0.047062158,
            0.046994456
          ],
          "watts": {
            "core": 5.784442098303887,
            "dram": 3.0502493353146,
            "package-0": 15.615531915211896,
            "psys": 27.899068604091685,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 5.561388,
            "dram": 2.8277509999999997,
            "package-0": 19.482433,
            "psys": 32.526284000000004,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.9679824633553276,
            "dram": 2.2401243293373336,
            "package-0": 15.562673359858774,
            "psys": 22.860813256404036,
            "uncore": 0.0
          }
        },
        "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": 1.637692,
            "dram": 0.148803,
            "package-0": 2.114252,
            "psys": 2.882805,
            "uncore": 0.0
          },
          "logits": {
            "l2": 992.2683369200568,
            "sha256": "ae79c6c31d292f024b23dc06077f4b134ca9d7fc1fefa9233a47dedcd490f0e2",
            "sum": 292038.3226532545,
            "top1": 78
          },
          "net_j": {
            "core": 1.5585142742876166,
            "dram": 0.12353456968295035,
            "package-0": 1.945699110433572,
            "psys": 2.467181805616026,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 46682,
          "npu_utilization": 0.8996752269178387,
          "seconds": 0.051887613,
          "watts": {
            "core": 31.562292140900755,
            "dram": 2.8677942845434035,
            "package-0": 40.74675780518175,
            "psys": 55.55863593108436,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 848.6195445371723,
          "decode_tokens_per_s": 20.557439560377347,
          "end_to_end_ms": 1206.6655859999998,
          "handoff_estimate_ms": 3.149337500000002,
          "handoff_plus_first_decode_ms": 51.887613,
          "output_tokens_per_s_e2e": 13.259680383393317,
          "prefill_tokens_per_s": 2113.1954394902805,
          "steady_tpot_ms": 48.7382755,
          "steady_tpot_p95_ms": 49.52061395,
          "ttft_ms": 477.00273300000003
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 47,
          "gross_j": {
            "core": 0.479979,
            "dram": 0.657774,
            "package-0": 7.415692,
            "psys": 11.833588,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1147.1172002671653,
            "sha256": "d036e9984fc486f880661b10e94e4628553614043ee547042d906b12c17261b9",
            "sum": 371321.3503918741,
            "top1": 291
          },
          "net_j": {
            "core": 0.0,
            "dram": 0.42548136553534,
            "package-0": 5.866185589322882,
            "psys": 8.012764721439359,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 467645,
          "npu_utilization": 0.9803822235123335,
          "seconds": 0.477002733,
          "watts": {
            "core": 1.0062395177094299,
            "dram": 1.3789732311659522,
            "package-0": 15.546435034786267,
            "psys": 24.808218446832253,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 953.3553678944744,
            "sha256": "d7f745389327c7b4f7c74a1b3a7960947e9141430a397b0cdfdaccad2789a89f",
            "sum": 273382.66879007034,
            "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": 3.443717,
            "dram": 2.021174,
            "package-0": 9.952489,
            "psys": 17.809891,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.409468189067711,
            "dram": 1.6911083941190432,
            "package-0": 7.75078866010232,
            "psys": 12.38086672934865,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 656514,
          "npu_utilization": 0.968630839922686,
          "seconds": 0.67777524,
          "token_latencies_s": [
            0.04844395,
            0.04877334,
            0.048878987,
            0.049748099,
            0.048703211,
            0.04881916,
            0.049398122,
            0.04914311,
            0.049133506,
            0.0480643,
            0.047195861,
            0.047228022,
            0.047117439,
            0.047075034
          ],
          "watts": {
            "core": 5.080912958697045,
            "dram": 2.9820711656566266,
            "package-0": 14.684055144888443,
            "psys": 26.27698674858645,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 10.58211692981422,
    "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": 406911,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.04545198939740658,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 9.380956999999999,
            "dram": 3.708669,
            "package-0": 32.590065,
            "psys": 53.031236,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.637229988590557,
            "dram": 2.8330480577081034,
            "package-0": 26.749241563943755,
            "psys": 38.62874369879418,
            "uncore": 0.0
          }
        },
        "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": 2.615655,
            "dram": 0.210998,
            "package-0": 3.276053,
            "psys": 4.890185,
            "uncore": 0.0
          },
          "logits": {
            "l2": 717.7345947887626,
            "sha256": "f0695fbadace6f6c8d0679619c9f0011a90337bca31c052537c786abd645c276",
            "sum": 148310.24642302655,
            "top1": 220
          },
          "net_j": {
            "core": 2.4041826100090113,
            "dram": 0.1435096337658141,
            "package-0": 2.8258723200701525,
            "psys": 3.780114845025571,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 52197,
          "npu_utilization": 0.37664412716470347,
          "seconds": 0.138584399,
          "watts": {
            "core": 18.87409419006825,
            "dram": 1.5225234696150756,
            "package-0": 23.639406914771122,
            "psys": 35.286691974613966,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1139.0122221389486,
          "decode_tokens_per_s": 17.979076763117355,
          "end_to_end_ms": 1798.0491869999998,
          "handoff_estimate_ms": 89.11109049999999,
          "handoff_plus_first_decode_ms": 138.584399,
          "output_tokens_per_s_e2e": 8.898532985460536,
          "prefill_tokens_per_s": 2108.4392336886735,
          "steady_tpot_ms": 49.4733085,
          "steady_tpot_p95_ms": 50.79916185,
          "ttft_ms": 963.7460580000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 96,
          "gross_j": {
            "core": 2.64721,
            "dram": 1.408382,
            "package-0": 18.469252,
            "psys": 28.88341,
            "uncore": 0.0
          },
          "logits": {
            "l2": 831.4911202378499,
            "sha256": "f586669e9cb29136fdd87403e7a53cdac5ca8787a455ff4d27734f7920186588",
            "sum": 207554.2427772805,
            "top1": 220
          },
          "net_j": {
            "core": 1.1765849967508688,
            "dram": 0.9390532199325051,
            "package-0": 15.338597620424776,
            "psys": 21.163740660018558,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 927824,
          "npu_utilization": 0.9627266356092322,
          "seconds": 0.963746058,
          "watts": {
            "core": 2.746792039278048,
            "dram": 1.4613621381992723,
            "package-0": 19.164023392560534,
            "psys": 29.96993840881682,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 674.5225862907852,
            "sha256": "4951873d057936f5eea807776d0b62282cdeffba732c9441debea531ffc8e8bb",
            "sum": 113242.55733524822,
            "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": 69,
          "gross_j": {
            "core": 4.118092,
            "dram": 2.089289,
            "package-0": 10.84476,
            "psys": 19.257641,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.056462381830676,
            "dram": 1.750485204009784,
            "package-0": 8.584771623448827,
            "psys": 13.68488819375005,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 677582,
          "npu_utilization": 0.9739309447655665,
          "seconds": 0.69571873,
          "token_latencies_s": [
            0.052205691,
            0.049185217,
            0.049444304,
            0.049973793,
            0.049305569,
            0.049236365,
            0.049572606,
            0.049463416,
            0.04962484,
            0.049378277,
            0.049483201,
            0.0500418,
            0.049516981,
            0.049259732
          ],
          "watts": {
            "core": 5.919190935106778,
            "dram": 3.00306562107362,
            "package-0": 15.587851142084391,
            "psys": 27.68021065064613,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 10.818759,
            "dram": 3.6581930000000003,
            "package-0": 33.903721,
            "psys": 54.00206,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.19865162213459,
            "dram": 2.8220234825617743,
            "package-0": 28.326058020417072,
            "psys": 40.24847730714521,
            "uncore": 0.0
          }
        },
        "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": 8,
          "gross_j": {
            "core": 2.383356,
            "dram": 0.174499,
            "package-0": 2.931511,
            "psys": 3.992605,
            "uncore": 0.0
          },
          "logits": {
            "l2": 717.7345947887626,
            "sha256": "f0695fbadace6f6c8d0679619c9f0011a90337bca31c052537c786abd645c276",
            "sum": 148310.24642302655,
            "top1": 220
          },
          "net_j": {
            "core": 2.2663090105534835,
            "dram": 0.13714513980713028,
            "package-0": 2.6823423340809174,
            "psys": 3.3781968036405523,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 48582,
          "npu_utilization": 0.6333656013173566,
          "seconds": 0.076704513,
          "watts": {
            "core": 31.071913591316328,
            "dram": 2.2749508884829237,
            "package-0": 38.21823365204079,
            "psys": 52.05176128293781,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1192.7521087911155,
          "decode_tokens_per_s": 19.655706813149937,
          "end_to_end_ms": 1717.037417,
          "handoff_estimate_ms": 27.8019725,
          "handoff_plus_first_decode_ms": 76.704513,
          "output_tokens_per_s_e2e": 9.31837584993059,
          "prefill_tokens_per_s": 2130.201722104344,
          "steady_tpot_ms": 48.9025405,
          "steady_tpot_p95_ms": 49.7028969,
          "ttft_ms": 953.900271
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 95,
          "gross_j": {
            "core": 4.054189,
            "dram": 1.402645,
            "package-0": 19.880686,
            "psys": 30.574385,
            "uncore": 0.0
          },
          "logits": {
            "l2": 831.4911202378499,
            "sha256": "f586669e9cb29136fdd87403e7a53cdac5ca8787a455ff4d27734f7920186588",
            "sum": 207554.2427772805,
            "top1": 220
          },
          "net_j": {
            "core": 2.5985881417719274,
            "dram": 0.9381109593383439,
            "package-0": 16.782014896450704,
            "psys": 22.933581061750647,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 926078,
          "npu_utilization": 0.9708331448833397,
          "seconds": 0.953900271,
          "watts": {
            "core": 4.250118301937247,
            "dram": 1.4704314933568143,
            "package-0": 20.841472221366,
            "psys": 32.05197223389823,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 674.5225862907852,
            "sha256": "4951873d057936f5eea807776d0b62282cdeffba732c9441debea531ffc8e8bb",
            "sum": 113242.55733524822,
            "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": 4.381214,
            "dram": 2.081049,
            "package-0": 11.091524,
            "psys": 19.43507,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.33375446980918,
            "dram": 1.7467673834163,
            "package-0": 8.86170078988545,
            "psys": 13.936699441754008,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 675317,
          "npu_utilization": 0.9838066658465521,
          "seconds": 0.686432633,
          "token_latencies_s": [
            0.049429486,
            0.05021066,
            0.04888858,
            0.049085281,
            0.048875295,
            0.048997466,
            0.048916501,
            0.048931819,
            0.04885273,
            0.048776068,
            0.048819037,
            0.048731533,
            0.049053754,
            0.048833118
          ],
          "watts": {
            "core": 6.382584086732951,
            "dram": 3.031687160479155,
            "package-0": 16.15821198873568,
            "psys": 28.313149849913966,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 9.619179,
            "dram": 3.66747,
            "package-0": 32.736976999999996,
            "psys": 53.363144000000005,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.974752667915164,
            "dram": 2.8235394387259185,
            "package-0": 27.10754403121824,
            "psys": 39.48190518484798,
            "uncore": 0.0
          }
        },
        "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": 9,
          "gross_j": {
            "core": 2.370417,
            "dram": 0.178283,
            "package-0": 2.949578,
            "psys": 4.105458,
            "uncore": 0.0
          },
          "logits": {
            "l2": 717.7345947887626,
            "sha256": "f0695fbadace6f6c8d0679619c9f0011a90337bca31c052537c786abd645c276",
            "sum": 148310.24642302655,
            "top1": 220
          },
          "net_j": {
            "core": 2.2323983376598506,
            "dram": 0.1342363325766008,
            "package-0": 2.655765009765463,
            "psys": 3.3809643774527642,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 48844,
          "npu_utilization": 0.5400235996637136,
          "seconds": 0.090447899,
          "watts": {
            "core": 26.207540763329398,
            "dram": 1.971112673385592,
            "package-0": 32.610796188864484,
            "psys": 45.39030807116924,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1181.783195202228,
          "decode_tokens_per_s": 19.106795571571624,
          "end_to_end_ms": 1732.9743799999999,
          "handoff_estimate_ms": 41.158146499999994,
          "handoff_plus_first_decode_ms": 90.44789899999999,
          "output_tokens_per_s_e2e": 9.232681212517406,
          "prefill_tokens_per_s": 2143.655779721865,
          "steady_tpot_ms": 49.289752500000006,
          "steady_tpot_p95_ms": 51.320964800000006,
          "ttft_ms": 947.913382
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 94,
          "gross_j": {
            "core": 2.506098,
            "dram": 1.398739,
            "package-0": 18.29402,
            "psys": 28.803087,
            "uncore": 0.0
          },
          "logits": {
            "l2": 831.4911202378499,
            "sha256": "f586669e9cb29136fdd87403e7a53cdac5ca8787a455ff4d27734f7920186588",
            "sum": 207554.2427772805,
            "top1": 220
          },
          "net_j": {
            "core": 1.0596328145790865,
            "dram": 0.9371204776820459,
            "package-0": 15.214796841386299,
            "psys": 21.2102384365836,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 926241,
          "npu_utilization": 0.9771367485557874,
          "seconds": 0.947913382,
          "watts": {
            "core": 2.643804853469196,
            "dram": 1.4755979043663294,
            "package-0": 19.299252808733954,
            "psys": 30.385779488868955,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 674.5225862907852,
            "sha256": "4951873d057936f5eea807776d0b62282cdeffba732c9441debea531ffc8e8bb",
            "sum": 113242.55733524822,
            "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": 69,
          "gross_j": {
            "core": 4.742664,
            "dram": 2.090448,
            "package-0": 11.493379,
            "psys": 20.454599,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.6827215156762265,
            "dram": 1.752182628467272,
            "package-0": 9.236982180066477,
            "psys": 14.890702370811617,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 677950,
          "npu_utilization": 0.9760109634788215,
          "seconds": 0.694613099,
          "token_latencies_s": [
            0.053529826,
            0.049208487,
            0.049084497,
            0.04905678,
            0.049075252,
            0.04901029,
            0.04950561,
            0.049404834,
            0.049118912,
            0.050131578,
            0.049157262,
            0.049469716,
            0.049371018,
            0.049445467
          ],
          "watts": {
            "core": 6.8277779483683485,
            "dram": 3.0095142216717683,
            "package-0": 16.546447247462574,
            "psys": 29.447470871838544,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 15.732889357022941,
    "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": 407089,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.046084214001894,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.506379,
            "dram": 5.885849,
            "package-0": 55.531901000000005,
            "psys": 94.529482,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.709011966885171,
            "dram": 4.35483843495905,
            "package-0": 45.31930562084578,
            "psys": 69.34693232462192,
            "uncore": 0.0
          }
        },
        "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": 17,
          "gross_j": {
            "core": 3.042351,
            "dram": 0.251098,
            "package-0": 3.700552,
            "psys": 5.670395,
            "uncore": 0.0
          },
          "logits": {
            "l2": 987.7546651143514,
            "sha256": "65b8cf3d23c292957632e47bc3a72984a0e7f6d6be836d09ac1b62daa0526b20",
            "sum": 285645.4547156822,
            "top1": 78
          },
          "net_j": {
            "core": 2.777340133277315,
            "dram": 0.16652360033401178,
            "package-0": 3.1363990523026652,
            "psys": 4.279288322171777,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 67409,
          "npu_utilization": 0.38814458602044144,
          "seconds": 0.173669819,
          "watts": {
            "core": 17.518017911909034,
            "dram": 1.4458355599483868,
            "package-0": 21.30797407003689,
            "psys": 32.650434212751726,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1302.8557450935916,
          "decode_tokens_per_s": 16.62785011452382,
          "end_to_end_ms": 3143.8630219999995,
          "handoff_estimate_ms": 122.5199525,
          "handoff_plus_first_decode_ms": 173.669819,
          "output_tokens_per_s_e2e": 5.089280254271842,
          "prefill_tokens_per_s": 1819.9969204135443,
          "steady_tpot_ms": 51.1498665,
          "steady_tpot_p95_ms": 56.3042718,
          "ttft_ms": 2241.7620349999997
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 222,
          "gross_j": {
            "core": 4.322498,
            "dram": 3.457999,
            "package-0": 40.939348,
            "psys": 68.50898,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1191.9284956676295,
            "sha256": "6ade4f6edb2f1529fce4ee1de9d91b6bcdcc6c2eb70a2a074a1cca9c1c183747",
            "sum": 385704.7659149729,
            "top1": 220
          },
          "net_j": {
            "core": 0.9016888847365134,
            "dram": 2.366297060101456,
            "package-0": 33.657157764739694,
            "psys": 50.55231858958943,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2163402,
          "npu_utilization": 0.9650453376511037,
          "seconds": 2.241762035,
          "watts": {
            "core": 1.9281698648268886,
            "dram": 1.5425361595081166,
            "package-0": 18.26212923621039,
            "psys": 30.560326622713994,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 816.3258561155944,
            "sha256": "7873c6e1ae7131c1ddafce3cc8875eb4cf52e122938a93b820e945a8560dabbe",
            "sum": 204753.52091340534,
            "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": 4.14153,
            "dram": 2.176752,
            "package-0": 10.892001,
            "psys": 20.350107,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.029982948871343,
            "dram": 1.8220177745235826,
            "package-0": 8.52574880380342,
            "psys": 14.515325412860715,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 711475,
          "npu_utilization": 0.9767223469493277,
          "seconds": 0.728431168,
          "token_latencies_s": [
            0.060425537,
            0.054085129,
            0.05262576,
            0.05130668,
            0.051174891,
            0.050962457,
            0.051116808,
            0.051124842,
            0.051195184,
            0.05118923,
            0.05072121,
            0.050913282,
            0.050618723,
            0.050939633
          ],
          "watts": {
            "core": 5.685547491564777,
            "dram": 2.988274109654792,
            "package-0": 14.952683902729435,
            "psys": 27.936897669925077,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.576502999999999,
            "dram": 5.826462,
            "package-0": 57.959874,
            "psys": 95.660217,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.916219808408012,
            "dram": 4.339199767788111,
            "package-0": 48.03910157223333,
            "psys": 71.19725384468838,
            "uncore": 0.0
          }
        },
        "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": 2.611748,
            "dram": 0.21881,
            "package-0": 3.281608,
            "psys": 4.7531,
            "uncore": 0.0
          },
          "logits": {
            "l2": 987.7546651143514,
            "sha256": "65b8cf3d23c292957632e47bc3a72984a0e7f6d6be836d09ac1b62daa0526b20",
            "sum": 285645.4547156822,
            "top1": 78
          },
          "net_j": {
            "core": 2.4149477374649306,
            "dram": 0.1560040318378363,
            "package-0": 2.8626612240062856,
            "psys": 3.7200474206981724,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 50876,
          "npu_utilization": 0.39448147737484807,
          "seconds": 0.1289693,
          "watts": {
            "core": 20.2509279340122,
            "dram": 1.6966053161488819,
            "package-0": 25.44487719170376,
            "psys": 36.85450723544285,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1341.1796973396185,
          "decode_tokens_per_s": 17.818719397930845,
          "end_to_end_ms": 3054.0277400000004,
          "handoff_estimate_ms": 78.1047915,
          "handoff_plus_first_decode_ms": 128.9693,
          "output_tokens_per_s_e2e": 5.238983192732885,
          "prefill_tokens_per_s": 1844.3041984803508,
          "steady_tpot_ms": 50.8645085,
          "steady_tpot_p95_ms": 51.4627742,
          "ttft_ms": 2212.2164030000004
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 219,
          "gross_j": {
            "core": 6.337508,
            "dram": 3.447623,
            "package-0": 43.023693,
            "psys": 69.963994,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1191.9284956676295,
            "sha256": "6ade4f6edb2f1529fce4ee1de9d91b6bcdcc6c2eb70a2a074a1cca9c1c183747",
            "sum": 385704.7659149729,
            "top1": 220
          },
          "net_j": {
            "core": 2.9617839409038678,
            "dram": 2.3703093061625182,
            "package-0": 35.83747946051998,
            "psys": 52.24399504603994,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2155902,
          "npu_utilization": 0.9745438995373004,
          "seconds": 2.212216403,
          "watts": {
            "core": 2.86477760105461,
            "dram": 1.55844744452878,
            "package-0": 19.448229812262177,
            "psys": 31.62619800898384,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 816.3258561155944,
            "sha256": "7873c6e1ae7131c1ddafce3cc8875eb4cf52e122938a93b820e945a8560dabbe",
            "sum": 204753.52091340534,
            "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": 71,
          "gross_j": {
            "core": 4.627247,
            "dram": 2.160029,
            "package-0": 11.654573,
            "psys": 20.943123,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.5394881300392136,
            "dram": 1.8128864297877565,
            "package-0": 9.33896088770706,
            "psys": 15.233211377950266,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 701125,
          "npu_utilization": 0.9835629264383575,
          "seconds": 0.712842037,
          "token_latencies_s": [
            0.051834426,
            0.050896155,
            0.050847052,
            0.050717886,
            0.050742519,
            0.050920484,
            0.050881965,
            0.050761774,
            0.05097453,
            0.050661685,
            0.050527041,
            0.050800322,
            0.051262654,
            0.050970986
          ],
          "watts": {
            "core": 6.491265609802975,
            "dram": 3.0301650125608406,
            "package-0": 16.349446855082142,
            "psys": 29.37975303496306,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 13.644496,
            "dram": 5.817246,
            "package-0": 58.016515000000005,
            "psys": 94.99444700000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.003646134877586,
            "dram": 4.336185633922896,
            "package-0": 48.13711207736704,
            "psys": 70.63349411528053,
            "uncore": 0.0
          }
        },
        "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": 12,
          "gross_j": {
            "core": 2.738701,
            "dram": 0.216369,
            "package-0": 3.400199,
            "psys": 4.85149,
            "uncore": 0.0
          },
          "logits": {
            "l2": 987.7546651143514,
            "sha256": "65b8cf3d23c292957632e47bc3a72984a0e7f6d6be836d09ac1b62daa0526b20",
            "sum": 285645.4547156822,
            "top1": 78
          },
          "net_j": {
            "core": 2.553586976575994,
            "dram": 0.1572925266672149,
            "package-0": 3.006129793456679,
            "psys": 3.8797813401651036,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 50762,
          "npu_utilization": 0.41844533963159436,
          "seconds": 0.121310946,
          "watts": {
            "core": 22.575877035861215,
            "dram": 1.783590080980821,
            "package-0": 28.02878975158598,
            "psys": 39.99218669022662,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1346.7958201770846,
          "decode_tokens_per_s": 17.92484973051763,
          "end_to_end_ms": 3041.29248,
          "handoff_estimate_ms": 70.3499075,
          "handoff_plus_first_decode_ms": 121.310946,
          "output_tokens_per_s_e2e": 5.260921172566737,
          "prefill_tokens_per_s": 1850.7888738761737,
          "steady_tpot_ms": 50.9610385,
          "steady_tpot_p95_ms": 51.880345350000006,
          "ttft_ms": 2204.465381
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 218,
          "gross_j": {
            "core": 6.228561,
            "dram": 3.43682,
            "package-0": 42.895276,
            "psys": 69.235967,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1191.9284956676295,
            "sha256": "6ade4f6edb2f1529fce4ee1de9d91b6bcdcc6c2eb70a2a074a1cca9c1c183747",
            "sum": 385704.7659149729,
            "top1": 220
          },
          "net_j": {
            "core": 2.8646645864328053,
            "dram": 2.363280928812686,
            "package-0": 35.73424105478665,
            "psys": 51.578054242284516,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2150338,
          "npu_utilization": 0.9754464817336137,
          "seconds": 2.204465381,
          "watts": {
            "core": 2.8254292644752583,
            "dram": 1.5590265239007626,
            "package-0": 19.458357735943054,
            "psys": 31.407146420504393,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 816.3258561155944,
            "sha256": "7873c6e1ae7131c1ddafce3cc8875eb4cf52e122938a93b820e945a8560dabbe",
            "sum": 204753.52091340534,
            "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": 71,
          "gross_j": {
            "core": 4.677234,
            "dram": 2.164057,
            "package-0": 11.72104,
            "psys": 20.90699,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.5853945718687874,
            "dram": 1.815612178442995,
            "package-0": 9.396741229123712,
            "psys": 15.175658532830905,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 701959,
          "npu_utilization": 0.981052624817542,
          "seconds": 0.715516153,
          "token_latencies_s": [
            0.05243529,
            0.051581529,
            0.051009574,
            0.050858536,
            0.050938977,
            0.05139146,
            0.050850895,
            0.050913271,
            0.05103691,
            0.0509831,
            0.050914064,
            0.050915901,
            0.050662856,
            0.050996186
          ],
          "watts": {
            "core": 6.536867099910182,
            "dram": 3.0244698053658055,
            "package-0": 16.381237447758917,
            "psys": 29.21945215679848,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 46.333827616646886,
    "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": 407278,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.04594239592552185,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 23.247012,
            "dram": 11.962311000000001,
            "package-0": 121.986687,
            "psys": 204.995446,
            "uncore": 0.0
          },
          "net_j": {
            "core": 13.409986591201342,
            "dram": 8.822966079028113,
            "package-0": 101.04570842965862,
            "psys": 153.35850029518318,
            "uncore": 0.0
          }
        },
        "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": 19,
          "gross_j": {
            "core": 3.354118,
            "dram": 0.301757,
            "package-0": 4.244923,
            "psys": 6.509322,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1201.451742992223,
            "sha256": "b76c1e342ce2c72df4501b64f4e3fc619680eb252044cc454aad186c16b9052a",
            "sum": 386158.7196010724,
            "top1": 261
          },
          "net_j": {
            "core": 3.050137738354793,
            "dram": 0.20474607944082163,
            "package-0": 3.5978123257482753,
            "psys": 4.913655463257111,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 57161,
          "npu_utilization": 0.28694176427514945,
          "seconds": 0.199207669,
          "watts": {
            "core": 16.837293548171584,
            "dram": 1.5147860597676086,
            "package-0": 21.309034041254705,
            "psys": 32.676061281556386,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1270.7656919998878,
          "decode_tokens_per_s": 15.173583703614803,
          "end_to_end_ms": 6446.5070559999995,
          "handoff_estimate_ms": 143.2111745,
          "handoff_plus_first_decode_ms": 199.207669,
          "output_tokens_per_s_e2e": 2.481964242187281,
          "prefill_tokens_per_s": 1497.999178598827,
          "steady_tpot_ms": 55.9964945,
          "steady_tpot_p95_ms": 58.5222284,
          "ttft_ms": 5457.946918000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 539,
          "gross_j": {
            "core": 15.216941,
            "dram": 9.326942,
            "package-0": 105.422948,
            "psys": 176.314123,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1277.917301221495,
            "sha256": "938fc863e80af83b1ea1b09a9e8a8e1594a85ece59922f15c0118b3801e46c64",
            "sum": 430820.31402779557,
            "top1": 220
          },
          "net_j": {
            "core": 6.888405555015245,
            "dram": 6.669009915676929,
            "package-0": 87.69323042503504,
            "psys": 132.59560904875454,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 5344647,
          "npu_utilization": 0.9792412935299273,
          "seconds": 5.457946918,
          "watts": {
            "core": 2.7880338941764693,
            "dram": 1.708873710229807,
            "package-0": 19.315495292253775,
            "psys": 32.304110986958484,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 1199.9686300410292,
            "sha256": "545756e65cf0318c97b46cc3463d1705c40d70fcc43ef1451157bbd1a8855e07",
            "sum": 387763.0401600674,
            "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": 78,
          "gross_j": {
            "core": 4.675953,
            "dram": 2.333612,
            "package-0": 12.318816,
            "psys": 22.172001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.4714432978313043,
            "dram": 1.9492100839103623,
            "package-0": 9.754665678875297,
            "psys": 15.849235783171533,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 766469,
          "npu_utilization": 0.9710098214692477,
          "seconds": 0.789352469,
          "token_latencies_s": [
            0.062525594,
            0.05636657,
            0.056303841,
            0.056092229,
            0.056279785,
            0.055745845,
            0.056270939,
            0.055863224,
            0.055574286,
            0.056256687,
            0.05556658,
            0.055216354,
            0.05590076,
            0.055360972
          ],
          "watts": {
            "core": 5.9237833333488945,
            "dram": 2.9563624510560693,
            "package-0": 15.606229769074176,
            "psys": 28.088847341022255,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 22.134894,
            "dram": 11.922271,
            "package-0": 120.69359899999999,
            "psys": 196.684921,
            "uncore": 0.0
          },
          "net_j": {
            "core": 12.34335123732851,
            "dram": 8.797441210196116,
            "package-0": 99.84944351166668,
            "psys": 145.28672479830834,
            "uncore": 0.0
          }
        },
        "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": 2.956048,
            "dram": 0.272948,
            "package-0": 3.745107,
            "psys": 5.522081,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1201.451742992223,
            "sha256": "b76c1e342ce2c72df4501b64f4e3fc619680eb252044cc454aad186c16b9052a",
            "sum": 386158.7196010724,
            "top1": 261
          },
          "net_j": {
            "core": 2.7212509168041037,
            "dram": 0.19801589400678496,
            "package-0": 3.2452729055199696,
            "psys": 4.289573841302266,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 55599,
          "npu_utilization": 0.36133798114187354,
          "seconds": 0.153869792,
          "watts": {
            "core": 19.211360212926003,
            "dram": 1.773889445434488,
            "package-0": 24.33945579129658,
            "psys": 35.88801237867404,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1276.6685193357519,
          "decode_tokens_per_s": 16.078689655995717,
          "end_to_end_ms": 6416.700871000001,
          "handoff_estimate_ms": 98.552389,
          "handoff_plus_first_decode_ms": 153.86979200000002,
          "output_tokens_per_s_e2e": 2.4934932018276403,
          "prefill_tokens_per_s": 1490.9399246130151,
          "steady_tpot_ms": 55.317403,
          "steady_tpot_p95_ms": 57.49918405,
          "ttft_ms": 5483.789028
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 538,
          "gross_j": {
            "core": 14.047267,
            "dram": 9.321265,
            "package-0": 104.173379,
            "psys": 169.149164,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1277.917301221495,
            "sha256": "938fc863e80af83b1ea1b09a9e8a8e1594a85ece59922f15c0118b3801e46c64",
            "sum": 430820.31402779557,
            "top1": 220
          },
          "net_j": {
            "core": 5.679297875585641,
            "dram": 6.650748225073016,
            "package-0": 86.35971533345065,
            "psys": 125.2236530472193,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 5359116,
          "npu_utilization": 0.9772651669560181,
          "seconds": 5.483789028,
          "watts": {
            "core": 2.5615987282288275,
            "dram": 1.6997854863500412,
            "package-0": 18.996605899332565,
            "psys": 30.845308442088374,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 1199.9686300410292,
            "sha256": "545756e65cf0318c97b46cc3463d1705c40d70fcc43ef1451157bbd1a8855e07",
            "sum": 387763.0401600674,
            "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": 77,
          "gross_j": {
            "core": 5.131579,
            "dram": 2.328058,
            "package-0": 12.775113,
            "psys": 22.013676,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.942802444938766,
            "dram": 1.9486770911163145,
            "package-0": 10.244455272696054,
            "psys": 15.77349790978678,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 763048,
          "npu_utilization": 0.9794695922004857,
          "seconds": 0.779042051,
          "token_latencies_s": [
            0.058245568,
            0.057097285,
            0.055307853,
            0.055386553,
            0.055481285,
            0.055178232,
            0.055225464,
            0.055286395,
            0.055123775,
            0.055317715,
            0.055381826,
            0.055428782,
            0.055317091,
            0.055220937
          ],
          "watts": {
            "core": 6.58703723812208,
            "dram": 2.988359867110691,
            "package-0": 16.398489636857867,
            "psys": 28.25736553212068,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 24.699887,
            "dram": 11.951142,
            "package-0": 123.271657,
            "psys": 196.55986099999998,
            "uncore": 0.0
          },
          "net_j": {
            "core": 14.86122190819071,
            "dram": 8.811273797820887,
            "package-0": 102.32718788598704,
            "psys": 144.91430819912776,
            "uncore": 0.0
          }
        },
        "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": 15,
          "gross_j": {
            "core": 2.917228,
            "dram": 0.271789,
            "package-0": 3.709524,
            "psys": 5.432542,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1201.451742992223,
            "sha256": "b76c1e342ce2c72df4501b64f4e3fc619680eb252044cc454aad186c16b9052a",
            "sum": 386158.7196010724,
            "top1": 261
          },
          "net_j": {
            "core": 2.6818414329577998,
            "dram": 0.19666876873090283,
            "package-0": 3.208435017169069,
            "psys": 4.1969404968129,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 55619,
          "npu_utilization": 0.36056272886817914,
          "seconds": 0.154256099,
          "watts": {
            "core": 18.91158935634694,
            "dram": 1.7619335751515406,
            "package-0": 24.04782711379211,
            "psys": 35.217680436739165,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 1270.5539099241466,
          "decode_tokens_per_s": 16.07177991201729,
          "end_to_end_ms": 6447.581591,
          "handoff_estimate_ms": 98.8833005,
          "handoff_plus_first_decode_ms": 154.256099,
          "output_tokens_per_s_e2e": 2.481550605320599,
          "prefill_tokens_per_s": 1482.6988858604689,
          "steady_tpot_ms": 55.3727985,
          "steady_tpot_p95_ms": 56.69758749999999,
          "ttft_ms": 5514.268661
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 545,
          "gross_j": {
            "core": 17.431168,
            "dram": 9.350196,
            "package-0": 107.573516,
            "psys": 170.200492,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1277.917301221495,
            "sha256": "938fc863e80af83b1ea1b09a9e8a8e1594a85ece59922f15c0118b3801e46c64",
            "sum": 430820.31402779557,
            "top1": 220
          },
          "net_j": {
            "core": 9.016688583784628,
            "dram": 6.664836135595192,
            "package-0": 89.66084160780647,
            "psys": 126.03083718130063,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 5352984,
          "npu_utilization": 0.9707513958939478,
          "seconds": 5.514268661,
          "watts": {
            "core": 3.16110241840101,
            "dram": 1.6956366428298697,
            "package-0": 19.50821090035388,
            "psys": 30.865469650355145,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 1199.9686300410292,
            "sha256": "545756e65cf0318c97b46cc3463d1705c40d70fcc43ef1451157bbd1a8855e07",
            "sum": 387763.0401600674,
            "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": 77,
          "gross_j": {
            "core": 4.351491,
            "dram": 2.329157,
            "package-0": 11.988617,
            "psys": 20.926827,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.162691891448282,
            "dram": 1.9497688934947917,
            "package-0": 9.457911261011498,
            "psys": 14.686530521014241,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 765561,
          "npu_utilization": 0.9826767053917278,
          "seconds": 0.779056831,
          "token_latencies_s": [
            0.056086977,
            0.056149778,
            0.055354418,
            0.055277658,
            0.055672377,
            0.055550868,
            0.057714948,
            0.055249627,
            0.055161305,
            0.055391179,
            0.055353252,
            0.055595278,
            0.055252188,
            0.055212434
          ],
          "watts": {
            "core": 5.585588659064078,
            "dram": 2.989713853109132,
            "package-0": 15.388629587666115,
            "psys": 26.861746367255716,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 11.477834473364055,
    "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": 407521,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "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": 0.06274531874805689,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 11.222139,
            "dram": 5.715562,
            "package-0": 40.526995,
            "psys": 69.437077,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.722327151526276,
            "dram": 4.59864747612006,
            "package-0": 33.076624397768775,
            "psys": 51.06571097353138,
            "uncore": 0.0
          }
        },
        "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": 3.173515,
            "dram": 0.348388,
            "package-0": 4.302723,
            "psys": 6.532759,
            "uncore": 0.0
          },
          "logits": {
            "l2": 656.7043815022596,
            "sha256": "2cabc748d175b793c08a889d7d9410edb1eb4a784b8bdcc4b1140fb6799d78e4",
            "sum": 94.01911521144211,
            "top1": 78
          },
          "net_j": {
            "core": 2.8972747869880875,
            "dram": 0.2602299162655138,
            "package-0": 3.714665112523682,
            "psys": 5.082706744566702,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 100319,
          "npu_utilization": 0.5541605746097014,
          "seconds": 0.181028757,
          "watts": {
            "core": 17.530446833924845,
            "dram": 1.9244898201449836,
            "package-0": 23.768174025522367,
            "psys": 36.08685773608886,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 446.4723727321237,
          "decode_tokens_per_s": 9.643070544697089,
          "end_to_end_ms": 2293.5349699999997,
          "handoff_estimate_ms": 83.11460550000002,
          "handoff_plus_first_decode_ms": 181.028757,
          "output_tokens_per_s_e2e": 6.976130823939433,
          "prefill_tokens_per_s": 1365.8280464846407,
          "steady_tpot_ms": 97.91415149999999,
          "steady_tpot_p95_ms": 99.5388777,
          "ttft_ms": 738.013839
        },
        "prefill": {
          "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": 2.111017,
            "dram": 1.308041,
            "package-0": 16.941301,
            "psys": 26.73571,
            "uncore": 0.0
          },
          "logits": {
            "l2": 690.669334639667,
            "sha256": "dffdd879fdeb67b8ccffdd2d86deb30dc4ae6c5324e8c93b70566bb7f425595c",
            "sum": 123939.967737481,
            "top1": 290
          },
          "net_j": {
            "core": 0.9848473048111885,
            "dram": 0.9486401683643302,
            "package-0": 14.543920239712133,
            "psys": 20.824170587601905,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 729906,
          "npu_utilization": 0.9890139743029942,
          "seconds": 0.738013839,
          "watts": {
            "core": 2.860403001196296,
            "dram": 1.772380043404579,
            "package-0": 22.955261954105445,
            "psys": 36.22657000067448,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 595.7391334219876,
            "sha256": "a4358e6f20367655c98e237fcca564f9503536c619fe29ca9486099ea9b451f3",
            "sum": -67153.35384979099,
            "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": 5.937607,
            "dram": 4.059133,
            "package-0": 19.282971,
            "psys": 36.168608,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.8402050597269994,
            "dram": 3.3897773914902167,
            "package-0": 14.818039045532958,
            "psys": 25.158833641362772,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1358163,
          "npu_utilization": 0.9881197056390508,
          "seconds": 1.374492374,
          "token_latencies_s": [
            0.099829065,
            0.097970373,
            0.097866055,
            0.097710669,
            0.099382623,
            0.098194411,
            0.097773425,
            0.098466244,
            0.098024195,
            0.097899874,
            0.097918433,
            0.097816606,
            0.097698954,
            0.09790987
          ],
          "watts": {
            "core": 4.319854451225933,
            "dram": 2.9531869923637717,
            "package-0": 14.029158229436637,
            "psys": 26.31415690924743,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 11.238434999999999,
            "dram": 5.659653,
            "package-0": 40.326923,
            "psys": 67.88605799999999,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.863117920013003,
            "dram": 4.58246918768118,
            "package-0": 33.141575834291395,
            "psys": 50.16819537864944,
            "uncore": 0.0
          }
        },
        "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": 10,
          "gross_j": {
            "core": 2.3034,
            "dram": 0.296264,
            "package-0": 3.268119,
            "psys": 4.716785,
            "uncore": 0.0
          },
          "logits": {
            "l2": 656.7043815022596,
            "sha256": "2cabc748d175b793c08a889d7d9410edb1eb4a784b8bdcc4b1140fb6799d78e4",
            "sum": 94.01911521144211,
            "top1": 78
          },
          "net_j": {
            "core": 2.1453308008945755,
            "dram": 0.24581849266910963,
            "package-0": 2.931622592444013,
            "psys": 3.8870412194756514,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 96983,
          "npu_utilization": 0.9362410780375372,
          "seconds": 0.103587636,
          "watts": {
            "core": 22.23624448771087,
            "dram": 2.860032446343307,
            "package-0": 31.549315402853676,
            "psys": 45.53424696360481,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 462.94000328706386,
          "decode_tokens_per_s": 10.186762902220105,
          "end_to_end_ms": 2211.949697,
          "handoff_estimate_ms": 5.935249500000003,
          "handoff_plus_first_decode_ms": 103.587636,
          "output_tokens_per_s_e2e": 7.233437551360373,
          "prefill_tokens_per_s": 1363.1743762424758,
          "steady_tpot_ms": 97.65238649999999,
          "steady_tpot_p95_ms": 98.4875902,
          "ttft_ms": 739.450519
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 74,
          "gross_j": {
            "core": 2.002619,
            "dram": 1.308102,
            "package-0": 16.780841,
            "psys": 26.173029,
            "uncore": 0.0
          },
          "logits": {
            "l2": 690.669334639667,
            "sha256": "dffdd879fdeb67b8ccffdd2d86deb30dc4ae6c5324e8c93b70566bb7f425595c",
            "sum": 123939.967737481,
            "top1": 290
          },
          "net_j": {
            "core": 0.87425700787557,
            "dram": 0.9480015283853644,
            "package-0": 14.378793296074114,
            "psys": 20.249981686267002,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 729099,
          "npu_utilization": 0.9860010660158858,
          "seconds": 0.739450519,
          "watts": {
            "core": 2.7082528831114394,
            "dram": 1.7690189761027135,
            "package-0": 22.693663157737262,
            "psys": 35.39524055699527,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 595.7391334219876,
            "sha256": "a4358e6f20367655c98e237fcca564f9503536c619fe29ca9486099ea9b451f3",
            "sum": -67153.35384979099,
            "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": 6.932416,
            "dram": 4.055287,
            "package-0": 20.277963,
            "psys": 36.996244,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.843530111242857,
            "dram": 3.3886491666267053,
            "package-0": 15.831159945773267,
            "psys": 26.03117247290679,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1355926,
          "npu_utilization": 0.9905139655839063,
          "seconds": 1.368911542,
          "token_latencies_s": [
            0.099556003,
            0.097912291,
            0.09766316,
            0.097610582,
            0.097817692,
            0.097761904,
            0.097874747,
            0.097827015,
            0.097549849,
            0.09738773,
            0.097427545,
            0.097396938,
            0.097447996,
            0.097641613
          ],
          "watts": {
            "core": 5.064181130266195,
            "dram": 2.9624171289221257,
            "package-0": 14.81320185990513,
            "psys": 27.02602970674638,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 11.764924,
            "dram": 5.662644,
            "package-0": 40.901323,
            "psys": 68.740912,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.394726083205628,
            "dram": 4.58709389485388,
            "package-0": 33.726873466681255,
            "psys": 51.049921114854904,
            "uncore": 0.0
          }
        },
        "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": 10,
          "gross_j": {
            "core": 2.355524,
            "dram": 0.29779,
            "package-0": 3.322196,
            "psys": 4.779956,
            "uncore": 0.0
          },
          "logits": {
            "l2": 656.7043815022596,
            "sha256": "2cabc748d175b793c08a889d7d9410edb1eb4a784b8bdcc4b1140fb6799d78e4",
            "sum": 94.01911521144211,
            "top1": 78
          },
          "net_j": {
            "core": 2.1981994597747634,
            "dram": 0.2475821398211931,
            "package-0": 2.9872848161274117,
            "psys": 3.9541211155663816,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 96896,
          "npu_utilization": 0.9398287121046924,
          "seconds": 0.103099638,
          "watts": {
            "core": 22.847063730718435,
            "dram": 2.888370956258838,
            "package-0": 32.22315872728865,
            "psys": 46.36249062290597,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 463.6431861516104,
          "decode_tokens_per_s": 10.188324121057455,
          "end_to_end_ms": 2208.594951,
          "handoff_estimate_ms": 5.349535999999988,
          "handoff_plus_first_decode_ms": 103.099638,
          "output_tokens_per_s_e2e": 7.244424783618912,
          "prefill_tokens_per_s": 1368.9673859858262,
          "steady_tpot_ms": 97.750102,
          "steady_tpot_p95_ms": 98.0997677,
          "ttft_ms": 736.321413
        },
        "prefill": {
          "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": 2.143489,
            "dram": 1.304501,
            "package-0": 16.955279,
            "psys": 26.389703,
            "uncore": 0.0
          },
          "logits": {
            "l2": 690.669334639667,
            "sha256": "dffdd879fdeb67b8ccffdd2d86deb30dc4ae6c5324e8c93b70566bb7f425595c",
            "sum": 123939.967737481,
            "top1": 290
          },
          "net_j": {
            "core": 1.0199018564814015,
            "dram": 0.9459243525187979,
            "package-0": 14.563395954410964,
            "psys": 20.491720031225135,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 730400,
          "npu_utilization": 0.9919581138135393,
          "seconds": 0.736321413,
          "watts": {
            "core": 2.9110779099398543,
            "dram": 1.771646154747913,
            "package-0": 23.027007908026167,
            "psys": 35.83992334608365,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 595.7391334219876,
            "sha256": "a4358e6f20367655c98e237fcca564f9503536c619fe29ca9486099ea9b451f3",
            "sum": -67153.35384979099,
            "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": 7.265911,
            "dram": 4.060353,
            "package-0": 20.623848,
            "psys": 37.571253,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.1766247669494625,
            "dram": 3.393587402513889,
            "package-0": 16.17619269614288,
            "psys": 26.604079968063388,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1354569,
          "npu_utilization": 0.989333056962304,
          "seconds": 1.3691739,
          "token_latencies_s": [
            0.098289257,
            0.097748221,
            0.097670707,
            0.097602527,
            0.097751983,
            0.097770435,
            0.097692025,
            0.097799931,
            0.097711887,
            0.097741773,
            0.097947507,
            0.097997735,
            0.097611626,
            0.097775414
          ],
          "watts": {
            "core": 5.306784623925419,
            "dram": 2.9655495185819714,
            "package-0": 15.062986520558125,
            "psys": 27.440818876258156,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 23.044777703471482,
    "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": 407644,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.05915387440472841,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 14.601586000000001,
            "dram": 7.2031670000000005,
            "package-0": 61.518457000000005,
            "psys": 98.975576,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.927391047926399,
            "dram": 5.7114650298800775,
            "package-0": 51.568069331241816,
            "psys": 74.43958662072386,
            "uncore": 0.0
          }
        },
        "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": 2.977104,
            "dram": 0.356505,
            "package-0": 4.123585,
            "psys": 6.31169,
            "uncore": 0.0
          },
          "logits": {
            "l2": 598.095176963933,
            "sha256": "0883b6dfe42b96f14222016a116c7843ca78c62813e9c5095fbc666e8cbfe150",
            "sum": 14349.924920428544,
            "top1": 220
          },
          "net_j": {
            "core": 2.7018475037117256,
            "dram": 0.2686608552838339,
            "package-0": 3.5376212405728547,
            "psys": 4.866801513585938,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 101607,
          "npu_utilization": 0.5632813628797887,
          "seconds": 0.180384097,
          "watts": {
            "core": 16.50424870879832,
            "dram": 1.9763660207806457,
            "package-0": 22.860025182818642,
            "psys": 34.9902796586331,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 668.5939786961023,
          "decode_tokens_per_s": 9.50839441818555,
          "end_to_end_ms": 3063.144547,
          "handoff_estimate_ms": 80.70304949999999,
          "handoff_plus_first_decode_ms": 180.384097,
          "output_tokens_per_s_e2e": 5.2233904585632995,
          "prefill_tokens_per_s": 1367.8056794497268,
          "steady_tpot_ms": 99.6810475,
          "steady_tpot_p95_ms": 100.4612143,
          "ttft_ms": 1485.591141
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 147,
          "gross_j": {
            "core": 5.465868,
            "dram": 2.71502,
            "package-0": 37.671961,
            "psys": 56.88053,
            "uncore": 0.0
          },
          "logits": {
            "l2": 599.2833854145028,
            "sha256": "c07e61c2839864d29446765c2cb7b6ae7bde6f2704446b2060059c8f9d6eff63",
            "sum": -5876.612534986809,
            "top1": 220
          },
          "net_j": {
            "core": 3.1989352759441734,
            "dram": 1.9915610845554839,
            "package-0": 32.846133311032375,
            "psys": 44.9808472072889,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1457431,
          "npu_utilization": 0.981044487798275,
          "seconds": 1.485591141,
          "watts": {
            "core": 3.6792545735839175,
            "dram": 1.827568787312794,
            "package-0": 25.358229434945184,
            "psys": 38.28814566147174,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 677.0484772729441,
            "sha256": "d7335174072bf9ed9b0a52e0cbb84640148dea91348cc0fb13fdba6b8c11a0d3",
            "sum": -131147.21031796,
            "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": 138,
          "gross_j": {
            "core": 6.158614,
            "dram": 4.131642,
            "package-0": 19.722911,
            "psys": 35.783356,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.0266082682705004,
            "dram": 3.45124309004076,
            "package-0": 15.184314779636585,
            "psys": 24.591937899849015,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1382592,
          "npu_utilization": 0.9895665407863609,
          "seconds": 1.397169309,
          "token_latencies_s": [
            0.101128282,
            0.099647102,
            0.099701659,
            0.099521264,
            0.099425763,
            0.099634718,
            0.099560782,
            0.099660436,
            0.099777722,
            0.099784135,
            0.100102024,
            0.09989721,
            0.099775444,
            0.099518445
          ],
          "watts": {
            "core": 4.40792247605834,
            "dram": 2.957151988227649,
            "package-0": 14.116335703162802,
            "psys": 25.611324103312377,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 15.065757,
            "dram": 7.170575,
            "package-0": 62.002099,
            "psys": 98.18059600000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 10.464868805558126,
            "dram": 5.702267825326905,
            "package-0": 52.20776615237241,
            "psys": 74.02941167167351,
            "uncore": 0.0
          }
        },
        "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": 2.486016,
            "dram": 0.322875,
            "package-0": 3.533927,
            "psys": 5.187365,
            "uncore": 0.0
          },
          "logits": {
            "l2": 598.095176963933,
            "sha256": "0883b6dfe42b96f14222016a116c7843ca78c62813e9c5095fbc666e8cbfe150",
            "sum": 14349.924920428544,
            "top1": 220
          },
          "net_j": {
            "core": 2.28264421951599,
            "dram": 0.2579718265711275,
            "package-0": 3.1009908307878558,
            "psys": 4.119816920135152,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 99110,
          "npu_utilization": 0.7436457771833315,
          "seconds": 0.133275819,
          "watts": {
            "core": 18.65316618313184,
            "dram": 2.42260750991896,
            "package-0": 26.515890328162232,
            "psys": 38.92202680817891,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 679.2468036896585,
          "decode_tokens_per_s": 9.842233963179135,
          "end_to_end_ms": 3015.1043609999997,
          "handoff_estimate_ms": 34.017079499999994,
          "handoff_plus_first_decode_ms": 133.27581899999998,
          "output_tokens_per_s_e2e": 5.306615653825457,
          "prefill_tokens_per_s": 1362.7887796789769,
          "steady_tpot_ms": 99.2587395,
          "steady_tpot_p95_ms": 99.68005465,
          "ttft_ms": 1491.060119
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 148,
          "gross_j": {
            "core": 5.133959,
            "dram": 2.718377,
            "package-0": 37.439906,
            "psys": 55.939127,
            "uncore": 0.0
          },
          "logits": {
            "l2": 599.2833854145028,
            "sha256": "c07e61c2839864d29446765c2cb7b6ae7bde6f2704446b2060059c8f9d6eff63",
            "sum": -5876.612534986809,
            "top1": 220
          },
          "net_j": {
            "core": 2.858680907657313,
            "dram": 1.9922547805122148,
            "package-0": 32.596312759842434,
            "psys": 43.995637333644986,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1453357,
          "npu_utilization": 0.9747138840885329,
          "seconds": 1.491060119,
          "watts": {
            "core": 3.4431602955373526,
            "dram": 1.8231169658156487,
            "package-0": 25.109588488698627,
            "psys": 37.516345777872694,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 677.0484772729441,
            "sha256": "d7335174072bf9ed9b0a52e0cbb84640148dea91348cc0fb13fdba6b8c11a0d3",
            "sum": -131147.21031796,
            "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": 7.445782,
            "dram": 4.129323,
            "package-0": 21.028266,
            "psys": 37.054104,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.323543678384823,
            "dram": 3.4520412182435622,
            "package-0": 16.510462561742116,
            "psys": 25.91395741789337,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1381352,
          "npu_utilization": 0.9932293379370176,
          "seconds": 1.390768423,
          "token_latencies_s": [
            0.100237533,
            0.099368385,
            0.099345932,
            0.099266446,
            0.099247309,
            0.09921427,
            0.099258712,
            0.099111658,
            0.099233088,
            0.099312333,
            0.099379874,
            0.099253428,
            0.099255143,
            0.099258767
          ],
          "watts": {
            "core": 5.35371804310803,
            "dram": 2.9690945895167196,
            "package-0": 15.119890308294696,
            "psys": 26.642899987671065,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 12.723234999999999,
            "dram": 7.186382999999999,
            "package-0": 59.613311,
            "psys": 94.738344,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.082295291720671,
            "dram": 5.705293961773884,
            "package-0": 49.73371681999548,
            "psys": 70.37691950663637,
            "uncore": 0.0
          }
        },
        "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": 2.485895,
            "dram": 0.322326,
            "package-0": 3.537893,
            "psys": 5.136217,
            "uncore": 0.0
          },
          "logits": {
            "l2": 598.095176963933,
            "sha256": "0883b6dfe42b96f14222016a116c7843ca78c62813e9c5095fbc666e8cbfe150",
            "sum": 14349.924920428544,
            "top1": 220
          },
          "net_j": {
            "core": 2.2811357678456523,
            "dram": 0.25698004136184305,
            "package-0": 3.1020032351135014,
            "psys": 4.061385847906301,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 99166,
          "npu_utilization": 0.7390241561841845,
          "seconds": 0.134185059,
          "watts": {
            "core": 18.525870305724577,
            "dram": 2.4021005199990264,
            "package-0": 26.36577444885276,
            "psys": 38.2771154872019,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 673.3848738937401,
          "decode_tokens_per_s": 9.815708810668962,
          "end_to_end_ms": 3041.351357,
          "handoff_estimate_ms": 34.76565049999999,
          "handoff_plus_first_decode_ms": 134.185059,
          "output_tokens_per_s_e2e": 5.260819327294844,
          "prefill_tokens_per_s": 1342.859649295615,
          "steady_tpot_ms": 99.4194085,
          "steady_tpot_p95_ms": 100.32548745,
          "ttft_ms": 1513.188665
        },
        "prefill": {
          "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": 4.185231,
            "dram": 2.726434,
            "package-0": 36.457182,
            "psys": 54.469954,
            "uncore": 0.0
          },
          "logits": {
            "l2": 599.2833854145028,
            "sha256": "c07e61c2839864d29446765c2cb7b6ae7bde6f2704446b2060059c8f9d6eff63",
            "sum": -5876.612534986809,
            "top1": 220
          },
          "net_j": {
            "core": 1.8761859285210432,
            "dram": 1.9895355356574238,
            "package-0": 31.541705893112688,
            "psys": 42.349213224105746,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1461733,
          "npu_utilization": 0.9659952085353481,
          "seconds": 1.513188665,
          "watts": {
            "core": 2.7658355476777245,
            "dram": 1.8017806127301383,
            "package-0": 24.09295208406812,
            "psys": 35.99680281771077,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 677.0484772729441,
            "sha256": "d7335174072bf9ed9b0a52e0cbb84640148dea91348cc0fb13fdba6b8c11a0d3",
            "sum": -131147.21031796,
            "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": 6.052109,
            "dram": 4.137623,
            "package-0": 19.618236,
            "psys": 35.132173,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.9249735953539755,
            "dram": 3.458778384754617,
            "package-0": 15.09000769176929,
            "psys": 23.96632043462432,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1382596,
          "npu_utilization": 0.9918351394379941,
          "seconds": 1.393977633,
          "token_latencies_s": [
            0.100106336,
            0.099676829,
            0.09966673,
            0.099440668,
            0.100732483,
            0.099398149,
            0.0992854,
            0.099314426,
            0.099307562,
            0.099366736,
            0.099546878,
            0.09923671,
            0.099396817,
            0.099462917
          ],
          "watts": {
            "core": 4.3416112688803805,
            "dram": 2.9682133357443905,
            "package-0": 14.073565841784205,
            "psys": 25.20282403986033,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 21.653414119966328,
    "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": 407815,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.05971842538565397,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 17.300981,
            "dram": 11.186616,
            "package-0": 105.502476,
            "psys": 173.102829,
            "uncore": 0.0
          },
          "net_j": {
            "core": 9.330816342289001,
            "dram": 8.643052788796465,
            "package-0": 88.53565520171955,
            "psys": 131.26549072083276,
            "uncore": 0.0
          }
        },
        "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": 21,
          "gross_j": {
            "core": 3.825246,
            "dram": 0.401305,
            "package-0": 5.096728,
            "psys": 7.693157,
            "uncore": 0.0
          },
          "logits": {
            "l2": 674.5130808654125,
            "sha256": "68592a61579d632d5b142b4d10b5be1b44d2f298fcb42b81fe8b59148d40b24f",
            "sum": -59216.823038564995,
            "top1": 1718
          },
          "net_j": {
            "core": 3.4874076937783944,
            "dram": 0.29348877236294374,
            "package-0": 4.377540599660544,
            "psys": 5.919761329011002,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 104775,
          "npu_utilization": 0.4732472611856444,
          "seconds": 0.221395893,
          "watts": {
            "core": 17.277854381878708,
            "dram": 1.8126126666676694,
            "package-0": 23.020878711602837,
            "psys": 34.7484178489255,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 784.209294141648,
          "decode_tokens_per_s": 9.085405482646086,
          "end_to_end_ms": 5223.095455000001,
          "handoff_estimate_ms": 119.53060350000001,
          "handoff_plus_first_decode_ms": 221.395893,
          "output_tokens_per_s_e2e": 3.0633175552408125,
          "prefill_tokens_per_s": 1142.1865657642632,
          "steady_tpot_ms": 101.8652895,
          "steady_tpot_p95_ms": 103.19750105,
          "ttft_ms": 3572.095945
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 353,
          "gross_j": {
            "core": 7.016889,
            "dram": 6.548994,
            "package-0": 79.992715,
            "psys": 127.320597,
            "uncore": 0.0
          },
          "logits": {
            "l2": 667.6443131718496,
            "sha256": "b618e00272bb302626d98d7f306a45e1f92c71ccf07eeff3434920f53951ad1a",
            "sum": -472.4365739785135,
            "top1": 36
          },
          "net_j": {
            "core": 1.5660614016765502,
            "dram": 4.809440911056631,
            "package-0": 68.38903815831776,
            "psys": 98.70787343326087,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3467063,
          "npu_utilization": 0.9705962699162606,
          "seconds": 3.572095945,
          "watts": {
            "core": 1.9643618503085867,
            "dram": 1.8333757269781286,
            "package-0": 22.39377559608075,
            "psys": 35.643106725119054,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 708.380856841874,
            "sha256": "fd2bd305053dee7b73acdaf34aa5d6eddb70fcf63291f691749f6eb56bcd0c8d",
            "sum": -134037.2010315787,
            "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": 141,
          "gross_j": {
            "core": 6.458846,
            "dram": 4.236317,
            "package-0": 20.413033,
            "psys": 38.089075,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.277347246834056,
            "dram": 3.540123105376891,
            "package-0": 15.769076443741241,
            "psys": 26.6378559585609,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1418238,
          "npu_utilization": 0.992049812364178,
          "seconds": 1.429603617,
          "token_latencies_s": [
            0.104826052,
            0.102320589,
            0.102107768,
            0.101992299,
            0.101964491,
            0.10183759,
            0.101881688,
            0.101848891,
            0.101732823,
            0.101914355,
            0.101827184,
            0.101836468,
            0.101734332,
            0.101752354
          ],
          "watts": {
            "core": 4.5179278530043065,
            "dram": 2.9632808350680047,
            "package-0": 14.278806207021509,
            "psys": 26.64310200888363,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 18.157363,
            "dram": 11.129732,
            "package-0": 106.278658,
            "psys": 172.197252,
            "uncore": 0.0
          },
          "net_j": {
            "core": 10.309671996053638,
            "dram": 8.625254490715564,
            "package-0": 89.57255810588867,
            "psys": 131.00280780180756,
            "uncore": 0.0
          }
        },
        "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": 2.80523,
            "dram": 0.374633,
            "package-0": 3.980703,
            "psys": 6.033615,
            "uncore": 0.0
          },
          "logits": {
            "l2": 674.5130808654125,
            "sha256": "68592a61579d632d5b142b4d10b5be1b44d2f298fcb42b81fe8b59148d40b24f",
            "sum": -59216.823038564995,
            "top1": 1718
          },
          "net_j": {
            "core": 2.5270395433290593,
            "dram": 0.28585252410211137,
            "package-0": 3.388493449904806,
            "psys": 4.573325439919474,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 102371,
          "npu_utilization": 0.5615314034453084,
          "seconds": 0.182306812,
          "watts": {
            "core": 15.387411853814873,
            "dram": 2.05495886791109,
            "package-0": 21.83518518221908,
            "psys": 33.09593829110456,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 796.4479229971188,
          "decode_tokens_per_s": 9.297816477719788,
          "end_to_end_ms": 5142.834681,
          "handoff_estimate_ms": 80.20840000000001,
          "handoff_plus_first_decode_ms": 182.306812,
          "output_tokens_per_s_e2e": 3.1111246992074952,
          "prefill_tokens_per_s": 1155.9538483692108,
          "steady_tpot_ms": 102.098412,
          "steady_tpot_p95_ms": 102.67646659999998,
          "ttft_ms": 3529.5526769999997
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 349,
          "gross_j": {
            "core": 9.22721,
            "dram": 6.530195,
            "package-0": 82.207309,
            "psys": 128.864294,
            "uncore": 0.0
          },
          "logits": {
            "l2": 667.6443131718496,
            "sha256": "b618e00272bb302626d98d7f306a45e1f92c71ccf07eeff3434920f53951ad1a",
            "sum": -472.4365739785135,
            "top1": 36
          },
          "net_j": {
            "core": 3.8413011560739125,
            "dram": 4.811359796131004,
            "package-0": 70.74183066733374,
            "psys": 100.59234480932811,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3465176,
          "npu_utilization": 0.9817606697246638,
          "seconds": 3.529552677,
          "watts": {
            "core": 2.6142717914732514,
            "dram": 1.8501480492282791,
            "package-0": 23.291140981035994,
            "psys": 36.51009229575525,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 708.380856841874,
            "sha256": "fd2bd305053dee7b73acdaf34aa5d6eddb70fcf63291f691749f6eb56bcd0c8d",
            "sum": -134037.2010315787,
            "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": 141,
          "gross_j": {
            "core": 6.124923,
            "dram": 4.224904,
            "package-0": 20.090646,
            "psys": 37.299343,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.941331296650666,
            "dram": 3.528042170482448,
            "package-0": 15.44223398865012,
            "psys": 25.837137552559966,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1417705,
          "npu_utilization": 0.9907264695613256,
          "seconds": 1.430975192,
          "token_latencies_s": [
            0.102719273,
            0.102653417,
            0.102406064,
            0.101976759,
            0.102039746,
            0.102068042,
            0.102182677,
            0.102017763,
            0.102527981,
            0.102028255,
            0.102100749,
            0.102096075,
            0.101958804,
            0.102174286
          ],
          "watts": {
            "core": 4.280244014181345,
            "dram": 2.9524648810263936,
            "package-0": 14.03982830192908,
            "psys": 26.06568108834133,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 21.815129,
            "dram": 11.166414,
            "package-0": 109.662866,
            "psys": 173.670759,
            "uncore": 0.0
          },
          "net_j": {
            "core": 13.921682586974912,
            "dram": 8.647334311181716,
            "package-0": 92.85936236835866,
            "psys": 132.2361334994205,
            "uncore": 0.0
          }
        },
        "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": 21,
          "gross_j": {
            "core": 2.947624,
            "dram": 0.393126,
            "package-0": 3.948232,
            "psys": 6.114304,
            "uncore": 0.0
          },
          "logits": {
            "l2": 674.5130808654125,
            "sha256": "68592a61579d632d5b142b4d10b5be1b44d2f298fcb42b81fe8b59148d40b24f",
            "sum": -59216.823038564995,
            "top1": 1718
          },
          "net_j": {
            "core": 2.634132078413583,
            "dram": 0.29307957022075376,
            "package-0": 3.280872982502247,
            "psys": 4.468708440251066,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 126162,
          "npu_utilization": 0.6141034567041583,
          "seconds": 0.205440954,
          "watts": {
            "core": 14.347791628732407,
            "dram": 1.9135717214397279,
            "package-0": 19.218329759119012,
            "psys": 29.76185556459205,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 791.831206974761,
          "decode_tokens_per_s": 9.131076469404587,
          "end_to_end_ms": 5172.819616,
          "handoff_estimate_ms": 103.144987,
          "handoff_plus_first_decode_ms": 205.440954,
          "output_tokens_per_s_e2e": 3.09309065224516,
          "prefill_tokens_per_s": 1155.7818514728679,
          "steady_tpot_ms": 102.295967,
          "steady_tpot_p95_ms": 104.3798861,
          "ttft_ms": 3530.077925
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 349,
          "gross_j": {
            "core": 11.187166,
            "dram": 6.532393,
            "package-0": 84.230863,
            "psys": 129.730258,
            "uncore": 0.0
          },
          "logits": {
            "l2": 667.6443131718496,
            "sha256": "b618e00272bb302626d98d7f306a45e1f92c71ccf07eeff3434920f53951ad1a",
            "sum": -472.4365739785135,
            "top1": 36
          },
          "net_j": {
            "core": 5.80045565568354,
            "dram": 4.813302008829955,
            "package-0": 72.76367843991221,
            "psys": 101.4541015382863,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3459794,
          "npu_utilization": 0.9800899791751764,
          "seconds": 3.530077925,
          "watts": {
            "core": 3.1690988804446856,
            "dram": 1.850495410805981,
            "package-0": 23.860907546396444,
            "psys": 36.749970044924716,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 708.380856841874,
            "sha256": "fd2bd305053dee7b73acdaf34aa5d6eddb70fcf63291f691749f6eb56bcd0c8d",
            "sum": -134037.2010315787,
            "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": 7.680339,
            "dram": 4.240895,
            "package-0": 21.483771,
            "psys": 37.826197,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.487094852877789,
            "dram": 3.540952732131008,
            "package-0": 16.814810945944195,
            "psys": 26.31332352088313,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1422136,
          "npu_utilization": 0.9894491552048791,
          "seconds": 1.437300737,
          "token_latencies_s": [
            0.107329655,
            0.102791549,
            0.102281906,
            0.102314533,
            0.102207795,
            0.102325921,
            0.102323002,
            0.102677839,
            0.102310028,
            0.102066061,
            0.102226658,
            0.102143562,
            0.102079772,
            0.102185705
          ],
          "watts": {
            "core": 5.3435852374435955,
            "dram": 2.950596831148776,
            "package-0": 14.947303961481238,
            "psys": 26.31752424962404,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 63.33386187534779,
    "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": 407992,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.06031969655305147,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 35.801361,
            "dram": 22.174321,
            "package-0": 226.071382,
            "psys": 366.207558,
            "uncore": 0.0
          },
          "net_j": {
            "core": 19.501787158602674,
            "dram": 16.972546863356182,
            "package-0": 191.37298361901188,
            "psys": 280.6471193020305,
            "uncore": 0.0
          }
        },
        "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": 28,
          "gross_j": {
            "core": 4.161,
            "dram": 0.492308,
            "package-0": 5.430772,
            "psys": 8.681374,
            "uncore": 0.0
          },
          "logits": {
            "l2": 648.2391451311615,
            "sha256": "e85afbe82578c61fa2818aef24e755aad5e884df578810e179b8993cd7997204",
            "sum": -34242.11205828376,
            "top1": 78
          },
          "net_j": {
            "core": 3.721802352290099,
            "dram": 0.35214439897812766,
            "package-0": 4.495811669348524,
            "psys": 6.375918423134816,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 135153,
          "npu_utilization": 0.4695750523491599,
          "seconds": 0.287819805,
          "watts": {
            "core": 14.45696205652005,
            "dram": 1.7104729815239783,
            "package-0": 18.86865290593884,
            "psys": 30.162531727099186,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 766.9252289948744,
          "decode_tokens_per_s": 8.293658544272171,
          "end_to_end_ms": 10681.614961,
          "handoff_estimate_ms": 179.63187399999998,
          "handoff_plus_first_decode_ms": 287.819805,
          "output_tokens_per_s_e2e": 1.497900837880614,
          "prefill_tokens_per_s": 921.4466493619024,
          "steady_tpot_ms": 108.187931,
          "steady_tpot_p95_ms": 110.89314859999999,
          "ttft_ms": 8873.004211
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 876,
          "gross_j": {
            "core": 24.734007,
            "dram": 17.223833,
            "package-0": 199.070169,
            "psys": 317.773905,
            "uncore": 0.0
          },
          "logits": {
            "l2": 607.3833136095392,
            "sha256": "597bcab04865e78eb1aaf108263fece075d0b3ccf1e95f89435d49f34b5a7e76",
            "sum": 46667.203135058284,
            "top1": 198
          },
          "net_j": {
            "core": 11.194276550973948,
            "dram": 12.902823116416771,
            "package-0": 170.24690247395776,
            "psys": 246.70055741782716,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 8685634,
          "npu_utilization": 0.9788831148341265,
          "seconds": 8.873004211,
          "watts": {
            "core": 2.787557225470137,
            "dram": 1.9411500987058419,
            "package-0": 22.43548681665333,
            "psys": 35.81356409208629,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 608.4144131591894,
            "sha256": "75dc1394dccec2f9374185c7ea194761337790f490c37e1fedd5be61da86a67d",
            "sum": -106184.73783594742,
            "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": 150,
          "gross_j": {
            "core": 6.906354,
            "dram": 4.45818,
            "package-0": 21.570441,
            "psys": 39.752279,
            "uncore": 0.0
          },
          "net_j": {
            "core": 4.585708255338626,
            "dram": 3.717579347961283,
            "package-0": 16.630269475705582,
            "psys": 27.57064346106854,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1506008,
          "npu_utilization": 0.9902794364678441,
          "seconds": 1.520790945,
          "token_latencies_s": [
            0.113353526,
            0.10956833,
            0.108401429,
            0.108262132,
            0.108255064,
            0.108037037,
            0.107998525,
            0.108458194,
            0.108389314,
            0.108120798,
            0.107970743,
            0.107897816,
            0.108096181,
            0.107928707
          ],
          "watts": {
            "core": 4.5412908478357625,
            "dram": 2.9314877331808415,
            "package-0": 14.183698996182542,
            "psys": 26.139213368343672,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 30.075605,
            "dram": 22.04041,
            "package-0": 220.912581,
            "psys": 362.20958299999995,
            "uncore": 0.0
          },
          "net_j": {
            "core": 13.829812997873127,
            "dram": 16.855799562435706,
            "package-0": 186.328672954945,
            "psys": 276.93145829420837,
            "uncore": 0.0
          }
        },
        "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": 3.225089,
            "dram": 0.466063,
            "package-0": 4.586292,
            "psys": 7.777019,
            "uncore": 0.0
          },
          "logits": {
            "l2": 648.2391451311615,
            "sha256": "e85afbe82578c61fa2818aef24e755aad5e884df578810e179b8993cd7997204",
            "sum": -34242.11205828376,
            "top1": 78
          },
          "net_j": {
            "core": 2.842729055237827,
            "dram": 0.3440383329193837,
            "package-0": 3.7723273026780038,
            "psys": 5.769918389105847,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 132923,
          "npu_utilization": 0.5304776306080121,
          "seconds": 0.2505723,
          "watts": {
            "core": 12.870891954138585,
            "dram": 1.8599941015028396,
            "package-0": 18.303268158531488,
            "psys": 31.037026039989254,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 769.4641418033639,
          "decode_tokens_per_s": 8.465157969907148,
          "end_to_end_ms": 10646.370058,
          "handoff_estimate_ms": 142.44708900000003,
          "handoff_plus_first_decode_ms": 250.5723,
          "output_tokens_per_s_e2e": 1.502859651959695,
          "prefill_tokens_per_s": 921.3016425952004,
          "steady_tpot_ms": 108.125211,
          "steady_tpot_p95_ms": 111.0312522,
          "ttft_ms": 8874.400763000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 876,
          "gross_j": {
            "core": 21.24494,
            "dram": 17.119768,
            "package-0": 195.062367,
            "psys": 312.728508,
            "uncore": 0.0
          },
          "logits": {
            "l2": 607.3833136095392,
            "sha256": "597bcab04865e78eb1aaf108263fece075d0b3ccf1e95f89435d49f34b5a7e76",
            "sum": 46667.203135058284,
            "top1": 198
          },
          "net_j": {
            "core": 7.7030784872227755,
            "dram": 12.798078018126336,
            "package-0": 166.23456388301784,
            "psys": 241.64397394441266,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 8671700,
          "npu_utilization": 0.9771589351874753,
          "seconds": 8.874400763,
          "watts": {
            "core": 2.3939576955524067,
            "dram": 1.9291181970705418,
            "package-0": 21.980342358807217,
            "psys": 35.239394337909275,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 608.4144131591894,
            "sha256": "75dc1394dccec2f9374185c7ea194761337790f490c37e1fedd5be61da86a67d",
            "sum": -106184.73783594742,
            "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": 151,
          "gross_j": {
            "core": 5.605576,
            "dram": 4.454579,
            "package-0": 21.263922,
            "psys": 41.704056,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.2840054554125246,
            "dram": 3.7136832113899856,
            "package-0": 16.321781769249156,
            "psys": 29.517565960689875,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1504336,
          "npu_utilization": 0.9887859677282983,
          "seconds": 1.521396995,
          "token_latencies_s": [
            0.115782487,
            0.108472895,
            0.108073991,
            0.108176431,
            0.108275055,
            0.107944352,
            0.108001312,
            0.108023482,
            0.107952397,
            0.10822568,
            0.107899985,
            0.10805141,
            0.10828673,
            0.108184362
          ],
          "watts": {
            "core": 3.6844926198897876,
            "dram": 2.9279530685546016,
            "package-0": 13.976576836869592,
            "psys": 27.411685534451845,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 24.925228999999998,
            "dram": 21.810308,
            "package-0": 216.495296,
            "psys": 359.33758099999994,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.763161781937455,
            "dram": 16.652417120908808,
            "package-0": 182.08962058483425,
            "psys": 274.49894810694786,
            "uncore": 0.0
          }
        },
        "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": 24,
          "gross_j": {
            "core": 3.270622,
            "dram": 0.463927,
            "package-0": 4.453419,
            "psys": 7.162579,
            "uncore": 0.0
          },
          "logits": {
            "l2": 648.2391451311615,
            "sha256": "e85afbe82578c61fa2818aef24e755aad5e884df578810e179b8993cd7997204",
            "sum": -34242.11205828376,
            "top1": 78
          },
          "net_j": {
            "core": 2.891443116728377,
            "dram": 0.34291752285929433,
            "package-0": 3.646226120146826,
            "psys": 5.172176556757655,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 134759,
          "npu_utilization": 0.5423166862231045,
          "seconds": 0.248487652,
          "watts": {
            "core": 13.162110767580515,
            "dram": 1.867002228344127,
            "package-0": 17.9220937706796,
            "psys": 28.82468783599758,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 773.4502172384254,
          "decode_tokens_per_s": 8.481422197261761,
          "end_to_end_ms": 10591.502617,
          "handoff_estimate_ms": 140.4493175,
          "handoff_plus_first_decode_ms": 248.487652,
          "output_tokens_per_s_e2e": 1.5106449555437995,
          "prefill_tokens_per_s": 926.6761486876783,
          "steady_tpot_ms": 108.0383345,
          "steady_tpot_p95_ms": 110.94088165,
          "ttft_ms": 8822.931303000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 872,
          "gross_j": {
            "core": 16.006245,
            "dram": 16.902179,
            "package-0": 191.791929,
            "psys": 314.443409,
            "uncore": 0.0
          },
          "logits": {
            "l2": 607.3833136095392,
            "sha256": "597bcab04865e78eb1aaf108263fece075d0b3ccf1e95f89435d49f34b5a7e76",
            "sum": 46667.203135058284,
            "top1": 198
          },
          "net_j": {
            "core": 2.5429231337940017,
            "dram": 12.605553814679343,
            "package-0": 163.13132043497043,
            "psys": 243.77114870575178,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 8631732,
          "npu_utilization": 0.978329276695718,
          "seconds": 8.822931303,
          "watts": {
            "core": 1.8141640743091252,
            "dram": 1.9157101443431694,
            "package-0": 21.737892137365538,
            "psys": 35.63933552254702,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 608.4144131591894,
            "sha256": "75dc1394dccec2f9374185c7ea194761337790f490c37e1fedd5be61da86a67d",
            "sum": -106184.73783594742,
            "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": 150,
          "gross_j": {
            "core": 5.648362,
            "dram": 4.444202,
            "package-0": 20.249948,
            "psys": 37.731593,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.3287955314150763,
            "dram": 3.7039457833701723,
            "package-0": 15.312074029717005,
            "psys": 25.555622844438425,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1505722,
          "npu_utilization": 0.9905520581800714,
          "seconds": 1.520083662,
          "token_latencies_s": [
            0.114859211,
            0.108831012,
            0.107915531,
            0.108096554,
            0.107850659,
            0.107825549,
            0.107884003,
            0.108296888,
            0.108025608,
            0.107923369,
            0.108051061,
            0.10799602,
            0.108304892,
            0.108178624
          ],
          "watts": {
            "core": 3.7158231097414425,
            "dram": 2.92365618491859,
            "package-0": 13.321600979091373,
            "psys": 24.82205022212784,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 235.41330679878592,
    "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": 408211,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 1008,
    "prompt_sha256": "2669029615323cd09063f0eb8d822f5daea575e1396736877003e47194c143f8",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "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": 0.046412899158895016,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 14.223962,
            "dram": 13.447170999999999,
            "package-0": 85.998559,
            "psys": 140.942816,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.514010024104772,
            "dram": 10.986650916884775,
            "package-0": 69.58567681295473,
            "psys": 100.47139755576656,
            "uncore": 0.0
          }
        },
        "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": 2.994072,
            "dram": 0.702269,
            "package-0": 5.266649,
            "psys": 8.406717,
            "uncore": 0.0
          },
          "logits": {
            "l2": 827.3468858325061,
            "sha256": "75e87e5dc1b7ba8c4b5a6f35fc368b20c39fd2db3eff946d358b57e684212379",
            "sum": 6878.080895855557,
            "top1": 78
          },
          "net_j": {
            "core": 2.566239245362325,
            "dram": 0.5657323408493733,
            "package-0": 4.355882160001381,
            "psys": 6.160918518864857,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 208200,
          "npu_utilization": 0.7425847482601501,
          "seconds": 0.280372039,
          "watts": {
            "core": 10.678925083538733,
            "dram": 2.5047754494520045,
            "package-0": 18.784501545819268,
            "psys": 29.98414902564517,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 202.66913529284705,
          "decode_tokens_per_s": 4.728974116777969,
          "end_to_end_ms": 5052.570035000001,
          "handoff_estimate_ms": 73.94627750000002,
          "handoff_plus_first_decode_ms": 280.37203900000003,
          "output_tokens_per_s_e2e": 3.166705238950735,
          "prefill_tokens_per_s": 535.9892533856925,
          "steady_tpot_ms": 206.4257615,
          "steady_tpot_p95_ms": 207.8030635,
          "ttft_ms": 1880.6347210000001
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 185,
          "gross_j": {
            "core": 3.985036,
            "dram": 3.734671,
            "package-0": 43.95289,
            "psys": 64.392231,
            "uncore": 0.0
          },
          "logits": {
            "l2": 852.4974085710298,
            "sha256": "7a1dee9f32cc78547b6aaf52f69c32ba16165afe1af665e77f2d4efbcabdb732",
            "sum": -209736.7540479228,
            "top1": 290
          },
          "net_j": {
            "core": 1.115287875249638,
            "dram": 2.8188322351793365,
            "package-0": 37.84379385708235,
            "psys": 49.32822313724799,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1871185,
          "npu_utilization": 0.9949752491036775,
          "seconds": 1.880634721,
          "watts": {
            "core": 2.1189845936062564,
            "dram": 1.9858566675904736,
            "package-0": 23.371306245281215,
            "psys": 34.2396268031042,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 764.6497230269619,
            "sha256": "76d71fd376731f9b77c6d5caa990ea4faf09a3dfc4e5c1add9c3562517997b0b",
            "sum": -165327.4534454923,
            "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": 284,
          "gross_j": {
            "core": 7.244854,
            "dram": 9.010231,
            "package-0": 36.77902,
            "psys": 68.143868,
            "uncore": 0.0
          },
          "net_j": {
            "core": 2.832482903492809,
            "dram": 7.602086340856065,
            "package-0": 27.38600079587099,
            "psys": 44.982255899653715,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2869046,
          "npu_utilization": 0.9922127676766817,
          "seconds": 2.891563275,
          "token_latencies_s": [
            0.209389031,
            0.206206506,
            0.206341476,
            0.206047133,
            0.206523493,
            0.205477716,
            0.20615996,
            0.205791867,
            0.206624987,
            0.206949081,
            0.206510047,
            0.206701205,
            0.206192714,
            0.206595171
          ],
          "watts": {
            "core": 2.5055145991920234,
            "dram": 3.1160414430149377,
            "package-0": 12.719424236012957,
            "psys": 23.566445385844094,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 14.958641,
            "dram": 13.44186,
            "package-0": 86.83181,
            "psys": 144.183043,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.362512600939299,
            "dram": 11.01766507138755,
            "package-0": 70.66123450468692,
            "psys": 104.30911227310797,
            "uncore": 0.0
          }
        },
        "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": 21,
          "gross_j": {
            "core": 2.545037,
            "dram": 0.650328,
            "package-0": 4.690845,
            "psys": 7.408246,
            "uncore": 0.0
          },
          "logits": {
            "l2": 827.3468858325061,
            "sha256": "75e87e5dc1b7ba8c4b5a6f35fc368b20c39fd2db3eff946d358b57e684212379",
            "sum": 6878.080895855557,
            "top1": 78
          },
          "net_j": {
            "core": 2.2199965773448094,
            "dram": 0.5465960325029529,
            "package-0": 3.9989016286604397,
            "psys": 5.702029541363593,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 204780,
          "npu_utilization": 0.9613676309443461,
          "seconds": 0.213009044,
          "watts": {
            "core": 11.948023202244878,
            "dram": 3.0530534656547257,
            "package-0": 22.021811430692118,
            "psys": 34.77902093208775,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 205.7060146978667,
          "decode_tokens_per_s": 4.8466610673536,
          "end_to_end_ms": 4977.977924,
          "handoff_estimate_ms": 7.175004000000013,
          "handoff_plus_first_decode_ms": 213.00904400000002,
          "output_tokens_per_s_e2e": 3.214156479654167,
          "prefill_tokens_per_s": 535.2978603368659,
          "steady_tpot_ms": 205.83404,
          "steady_tpot_p95_ms": 206.16559439999997,
          "ttft_ms": 1883.063757
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 187,
          "gross_j": {
            "core": 4.707264,
            "dram": 3.783742,
            "package-0": 44.781624,
            "psys": 67.045849,
            "uncore": 0.0
          },
          "logits": {
            "l2": 852.4974085710298,
            "sha256": "7a1dee9f32cc78547b6aaf52f69c32ba16165afe1af665e77f2d4efbcabdb732",
            "sum": -209736.7540479228,
            "top1": 290
          },
          "net_j": {
            "core": 1.8338092960675607,
            "dram": 2.866720333840175,
            "package-0": 38.66463732188544,
            "psys": 51.96238439905513,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1876379,
          "npu_utilization": 0.9964500633740359,
          "seconds": 1.883063757,
          "watts": {
            "core": 2.4997900270245603,
            "dram": 2.009354163359855,
            "package-0": 23.781257450009964,
            "psys": 35.604662216437106,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 764.6497230269619,
            "sha256": "76d71fd376731f9b77c6d5caa990ea4faf09a3dfc4e5c1add9c3562517997b0b",
            "sum": -165327.4534454923,
            "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": 282,
          "gross_j": {
            "core": 7.70634,
            "dram": 9.00779,
            "package-0": 37.359341,
            "psys": 69.728948,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.3087067275269293,
            "dram": 7.604348705044422,
            "package-0": 27.99769555414104,
            "psys": 46.64469833268925,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2865011,
          "npu_utilization": 0.9941378628792562,
          "seconds": 2.881905123,
          "token_latencies_s": [
            0.205722778,
            0.205592499,
            0.205914427,
            0.206225774,
            0.205656983,
            0.20613319,
            0.205656623,
            0.205998249,
            0.205715775,
            0.205775709,
            0.205675284,
            0.205892371,
            0.205952179,
            0.205928122
          ],
          "watts": {
            "core": 2.674043617361653,
            "dram": 3.125637248815148,
            "package-0": 12.963418088208867,
            "psys": 24.195434972339996,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 16.2524,
            "dram": 13.396022,
            "package-0": 88.18233000000001,
            "psys": 146.024651,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.640808787476434,
            "dram": 10.966892337269128,
            "package-0": 71.9788373948995,
            "psys": 106.06955219372988,
            "uncore": 0.0
          }
        },
        "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": 21,
          "gross_j": {
            "core": 2.636712,
            "dram": 0.649351,
            "package-0": 4.792102,
            "psys": 7.419964,
            "uncore": 0.0
          },
          "logits": {
            "l2": 827.3468858325061,
            "sha256": "75e87e5dc1b7ba8c4b5a6f35fc368b20c39fd2db3eff946d358b57e684212379",
            "sum": 6878.080895855557,
            "top1": 78
          },
          "net_j": {
            "core": 2.313368344491132,
            "dram": 0.5461605312895796,
            "package-0": 4.103770692569335,
            "psys": 5.722654285970559,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 204825,
          "npu_utilization": 0.9666248381879695,
          "seconds": 0.2118971,
          "watts": {
            "core": 12.443360480157587,
            "dram": 3.064463836456469,
            "package-0": 22.615231638375416,
            "psys": 35.01682656345934,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 205.28812655271398,
          "decode_tokens_per_s": 4.847297358948806,
          "end_to_end_ms": 4988.111184,
          "handoff_estimate_ms": 6.186430500000006,
          "handoff_plus_first_decode_ms": 211.8971,
          "output_tokens_per_s_e2e": 3.207626977386156,
          "prefill_tokens_per_s": 532.3184701415584,
          "steady_tpot_ms": 205.7106695,
          "steady_tpot_p95_ms": 206.5551994,
          "ttft_ms": 1893.603278
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 187,
          "gross_j": {
            "core": 3.676626,
            "dram": 3.740835,
            "package-0": 43.650828,
            "psys": 65.083696,
            "uncore": 0.0
          },
          "logits": {
            "l2": 852.4974085710298,
            "sha256": "7a1dee9f32cc78547b6aaf52f69c32ba16165afe1af665e77f2d4efbcabdb732",
            "sum": -209736.7540479228,
            "top1": 290
          },
          "net_j": {
            "core": 0.7870885500202931,
            "dram": 2.818680757169832,
            "package-0": 37.499604504724545,
            "psys": 49.91580914233469,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 1881441,
          "npu_utilization": 0.993577177362702,
          "seconds": 1.893603278,
          "watts": {
            "core": 1.9416031027804337,
            "dram": 1.9755114724722187,
            "package-0": 23.05172815612331,
            "psys": 34.37029115662526,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 764.6497230269619,
            "sha256": "76d71fd376731f9b77c6d5caa990ea4faf09a3dfc4e5c1add9c3562517997b0b",
            "sum": -165327.4534454923,
            "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": 9.939062,
            "dram": 9.005836,
            "package-0": 39.7394,
            "psys": 73.520991,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.54035189296501,
            "dram": 7.602051048809717,
            "package-0": 30.37546219760562,
            "psys": 50.43108876542463,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2864769,
          "npu_utilization": 0.9938105394030774,
          "seconds": 2.882610806,
          "token_latencies_s": [
            0.206436279,
            0.205604898,
            0.205548187,
            0.205792852,
            0.205689408,
            0.205449851,
            0.205719907,
            0.205701432,
            0.205562851,
            0.206477729,
            0.206443803,
            0.206699073,
            0.205846886,
            0.205566251
          ],
          "watts": {
            "core": 3.4479375361087157,
            "dram": 3.124194213542402,
            "package-0": 13.785905442831398,
            "psys": 25.50500083013981,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 383.0334553383291,
    "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": 408344,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 2032,
    "prompt_sha256": "cd1a97358e68d07c27508cfacc4ba6882a7a2962b1281e2e2b0c7e5e320da4fa",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.04666558373719454,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 18.804945,
            "dram": 17.360122999999998,
            "package-0": 133.89669,
            "psys": 213.593874,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.122277777820676,
            "dram": 13.950903678843732,
            "package-0": 111.15551611082272,
            "psys": 157.51794815752865,
            "uncore": 0.0
          }
        },
        "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": 3.278557,
            "dram": 0.718626,
            "package-0": 5.611619,
            "psys": 9.057289,
            "uncore": 0.0
          },
          "logits": {
            "l2": 656.4427383867821,
            "sha256": "bbd35af41712a9004a15d466ce75c6984ad14e8615b96488be7f0c044f186394",
            "sum": -33059.47061268799,
            "top1": 220
          },
          "net_j": {
            "core": 2.8291799701694265,
            "dram": 0.5752137958454586,
            "package-0": 4.654988884633149,
            "psys": 6.698399363545661,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 209202,
          "npu_utilization": 0.7103858356883966,
          "seconds": 0.294490669,
          "watts": {
            "core": 11.132974131686327,
            "dram": 2.440233513816358,
            "package-0": 19.055337199835016,
            "psys": 30.755775830710622,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 292.54291416283405,
          "decode_tokens_per_s": 4.67141247148557,
          "end_to_end_ms": 7000.682296,
          "handoff_estimate_ms": 86.46781199999998,
          "handoff_plus_first_decode_ms": 294.49066899999997,
          "output_tokens_per_s_e2e": 2.285491516897141,
          "prefill_tokens_per_s": 536.1955684990492,
          "steady_tpot_ms": 208.02285700000002,
          "steady_tpot_p95_ms": 209.51920255000002,
          "ttft_ms": 3789.662055
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 374,
          "gross_j": {
            "core": 7.47343,
            "dram": 7.555705,
            "package-0": 90.310254,
            "psys": 134.026085,
            "uncore": 0.0
          },
          "logits": {
            "l2": 639.6205341218227,
            "sha256": "0745c5d78f0230e7056bb6b8f550d6c0caf532bb86cab6ab11e739777c0918a6",
            "sum": -21376.843404488638,
            "top1": 220
          },
          "net_j": {
            "core": 1.6906081397341044,
            "dram": 5.7102007259464855,
            "package-0": 77.99983051124754,
            "psys": 103.67064257610521,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3753151,
          "npu_utilization": 0.990365617179023,
          "seconds": 3.789662055,
          "watts": {
            "core": 1.9720571099841777,
            "dram": 1.9937674891171793,
            "package-0": 23.83068798465725,
            "psys": 35.36623663399453,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 840.7172376253137,
            "sha256": "398e2c128d6fbbccb88ad985ccb32f4f56ee6a0eb59aed5f4d2565cdfaec3cbf",
            "sum": -213465.7986249458,
            "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": 287,
          "gross_j": {
            "core": 8.052958,
            "dram": 9.085792,
            "package-0": 37.974817,
            "psys": 70.5105,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.602489667917145,
            "dram": 7.665489157051788,
            "package-0": 28.50069671494203,
            "psys": 47.14890621787778,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2895714,
          "npu_utilization": 0.9928628969855684,
          "seconds": 2.916529572,
          "token_latencies_s": [
            0.211474853,
            0.20819779,
            0.207863707,
            0.207756131,
            0.207890492,
            0.207873048,
            0.208079224,
            0.20796649,
            0.207883153,
            0.207922069,
            0.208460489,
            0.208225288,
            0.208412243,
            0.20846616
          ],
          "watts": {
            "core": 2.7611439559235165,
            "dram": 3.115275115749795,
            "package-0": 13.020549273552849,
            "psys": 24.17616494512266,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 25.226559,
            "dram": 17.473343,
            "package-0": 140.648687,
            "psys": 223.515542,
            "uncore": 0.0
          },
          "net_j": {
            "core": 14.61191351712577,
            "dram": 14.085831836841587,
            "package-0": 118.05231722852187,
            "psys": 167.79667885822093,
            "uncore": 0.0
          }
        },
        "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": 26,
          "gross_j": {
            "core": 2.763726,
            "dram": 0.699705,
            "package-0": 5.036608,
            "psys": 8.131938,
            "uncore": 0.0
          },
          "logits": {
            "l2": 656.4427383867821,
            "sha256": "bbd35af41712a9004a15d466ce75c6984ad14e8615b96488be7f0c044f186394",
            "sum": -33059.47061268799,
            "top1": 220
          },
          "net_j": {
            "core": 2.3627363837951996,
            "dram": 0.5717349417068658,
            "package-0": 4.1829846117384015,
            "psys": 6.027045699768429,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 207401,
          "npu_utilization": 0.7892544701688746,
          "seconds": 0.2627809,
          "watts": {
            "core": 10.517225567002779,
            "dram": 2.6626935214850094,
            "package-0": 19.166568042045675,
            "psys": 30.945696586015195,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 294.41761434710696,
          "decode_tokens_per_s": 4.7291906814509685,
          "end_to_end_ms": 6956.105546,
          "handoff_estimate_ms": 55.00022299999996,
          "handoff_plus_first_decode_ms": 262.7809,
          "output_tokens_per_s_e2e": 2.300137612086773,
          "prefill_tokens_per_s": 536.9531182202691,
          "steady_tpot_ms": 207.78067700000003,
          "steady_tpot_p95_ms": 208.00556804999997,
          "ttft_ms": 3784.315485
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 375,
          "gross_j": {
            "core": 9.476355,
            "dram": 7.695415,
            "package-0": 92.425239,
            "psys": 137.867567,
            "uncore": 0.0
          },
          "logits": {
            "l2": 639.6205341218227,
            "sha256": "0745c5d78f0230e7056bb6b8f550d6c0caf532bb86cab6ab11e739777c0918a6",
            "sum": -21376.843404488638,
            "top1": 220
          },
          "net_j": {
            "core": 3.7016917199517394,
            "dram": 5.852514419268825,
            "package-0": 80.13218342953604,
            "psys": 107.55495095361458,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3750046,
          "npu_utilization": 0.9909443371896887,
          "seconds": 3.784315485,
          "watts": {
            "core": 2.5041133693957867,
            "dram": 2.0335025001225553,
            "package-0": 24.423238328397456,
            "psys": 36.431309056147576,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 840.7172376253137,
            "sha256": "398e2c128d6fbbccb88ad985ccb32f4f56ee6a0eb59aed5f4d2565cdfaec3cbf",
            "sum": -213465.7986249458,
            "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": 288,
          "gross_j": {
            "core": 12.986478,
            "dram": 9.078223,
            "package-0": 43.18684,
            "psys": 77.516037,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.54748541337883,
            "dram": 7.661582475865895,
            "package-0": 33.73714918724743,
            "psys": 54.21468220483793,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2895085,
          "npu_utilization": 0.9952134351494398,
          "seconds": 2.909009161,
          "token_latencies_s": [
            0.208025889,
            0.207849307,
            0.207632777,
            0.207994626,
            0.20796089,
            0.207967161,
            0.207593927,
            0.207645205,
            0.207481571,
            0.207664276,
            0.207588158,
            0.207944997,
            0.207712047,
            0.207914855
          ],
          "watts": {
            "core": 4.464227261331749,
            "dram": 3.1207268515026856,
            "package-0": 14.845893433059558,
            "psys": 26.64688652040996,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 23.181887,
            "dram": 17.159014,
            "package-0": 137.77161999999998,
            "psys": 212.96588599999998,
            "uncore": 0.0
          },
          "net_j": {
            "core": 12.580058526906644,
            "dram": 13.77559320083736,
            "package-0": 115.20253497324937,
            "psys": 157.31430246873828,
            "uncore": 0.0
          }
        },
        "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": 26,
          "gross_j": {
            "core": 2.766839,
            "dram": 0.693419,
            "package-0": 5.011218,
            "psys": 7.827983,
            "uncore": 0.0
          },
          "logits": {
            "l2": 656.4427383867821,
            "sha256": "bbd35af41712a9004a15d466ce75c6984ad14e8615b96488be7f0c044f186394",
            "sum": -33059.47061268799,
            "top1": 220
          },
          "net_j": {
            "core": 2.3749461571720865,
            "dram": 0.5683520458578848,
            "package-0": 4.176959747887614,
            "psys": 5.770841882003985,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 207966,
          "npu_utilization": 0.8097749492179338,
          "seconds": 0.256819503,
          "watts": {
            "core": 10.773476966038674,
            "dram": 2.7000246939968573,
            "package-0": 19.51260687549886,
            "psys": 30.48048496534938,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 294.7735485571657,
          "decode_tokens_per_s": 4.728065618031263,
          "end_to_end_ms": 6947.706163,
          "handoff_estimate_ms": 48.7699865,
          "handoff_plus_first_decode_ms": 256.819503,
          "output_tokens_per_s_e2e": 2.3029183481028572,
          "prefill_tokens_per_s": 538.2551381755361,
          "steady_tpot_ms": 208.0495165,
          "steady_tpot_p95_ms": 209.2143584,
          "ttft_ms": 3775.161361
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 371,
          "gross_j": {
            "core": 8.595437,
            "dram": 7.393658,
            "package-0": 90.820934,
            "psys": 131.256928,
            "uncore": 0.0
          },
          "logits": {
            "l2": 639.6205341218227,
            "sha256": "0745c5d78f0230e7056bb6b8f550d6c0caf532bb86cab6ab11e739777c0918a6",
            "sum": -21376.843404488638,
            "top1": 220
          },
          "net_j": {
            "core": 2.83474242418196,
            "dram": 5.555215329953745,
            "package-0": 78.55761489190999,
            "psys": 101.01763708920457,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3728764,
          "npu_utilization": 0.9877098336830535,
          "seconds": 3.775161361,
          "watts": {
            "core": 2.276839630961671,
            "dram": 1.9585011852424499,
            "package-0": 24.057497233957307,
            "psys": 34.768560982842715,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 840.7172376253137,
            "sha256": "398e2c128d6fbbccb88ad985ccb32f4f56ee6a0eb59aed5f4d2565cdfaec3cbf",
            "sum": -213465.7986249458,
            "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": 287,
          "gross_j": {
            "core": 11.819611,
            "dram": 9.071937,
            "package-0": 41.939468,
            "psys": 73.880975,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.370369945552597,
            "dram": 7.65202582502573,
            "package-0": 32.46796033345177,
            "psys": 50.525823497529714,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2897016,
          "npu_utilization": 0.9935833121843073,
          "seconds": 2.915725299,
          "token_latencies_s": [
            0.209631232,
            0.208989888,
            0.208259828,
            0.208009894,
            0.208339283,
            0.207821217,
            0.2079403,
            0.208638078,
            0.208150652,
            0.2079932,
            0.207863392,
            0.208021276,
            0.207932833,
            0.208077757
          ],
          "watts": {
            "core": 4.053746422563795,
            "dram": 3.1113826131396474,
            "package-0": 14.383888638063361,
            "psys": 25.33879821440614,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 15.589648451656103,
    "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": 408602,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 4080,
    "prompt_sha256": "66f8de7a9ef70b33481de3e8dae7cef6e4434d9723c6203a6a3dbe07ef228a16",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.046240122988820076,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 39.246176,
            "dram": 26.771172,
            "package-0": 248.744115,
            "psys": 390.289466,
            "uncore": 0.0
          },
          "net_j": {
            "core": 21.47481823554713,
            "dram": 21.09969923073094,
            "package-0": 210.91259514050964,
            "psys": 297.0032745093148,
            "uncore": 0.0
          }
        },
        "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": 4.64397,
            "dram": 0.777647,
            "package-0": 7.254071,
            "psys": 11.181428,
            "uncore": 0.0
          },
          "logits": {
            "l2": 738.8544173581239,
            "sha256": "756a111ecb880182890767495a4326b190317a49fe7bcb1d25e5e6a9434fe988",
            "sum": -54558.6855568951,
            "top1": 78
          },
          "net_j": {
            "core": 4.1402585429150776,
            "dram": 0.6168947462729233,
            "package-0": 6.181774203956754,
            "psys": 8.537323702259368,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 215530,
          "npu_utilization": 0.6529279058896805,
          "seconds": 0.3300977,
          "watts": {
            "core": 14.068471243513665,
            "dram": 2.3558085984846304,
            "package-0": 21.975527245418554,
            "psys": 33.873086664947984,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 351.7051022920916,
          "decode_tokens_per_s": 4.503460418454397,
          "end_to_end_ms": 11646.120495,
          "handoff_estimate_ms": 115.8443975,
          "handoff_plus_first_decode_ms": 330.0977,
          "output_tokens_per_s_e2e": 1.373848055828483,
          "prefill_tokens_per_s": 490.6589323553983,
          "steady_tpot_ms": 214.2533025,
          "steady_tpot_p95_ms": 215.05255904999999,
          "ttft_ms": 8315.348465
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 821,
          "gross_j": {
            "core": 21.158271,
            "dram": 16.673908,
            "package-0": 196.955917,
            "psys": 297.429353,
            "uncore": 0.0
          },
          "logits": {
            "l2": 696.8009887540125,
            "sha256": "ee9a1e49729de1ec15cfac4c7e2d3b98754929839e7fba889ae61a8056468138",
            "sum": 28229.591541170143,
            "top1": 49
          },
          "net_j": {
            "core": 8.469493430589406,
            "dram": 12.624467466864697,
            "package-0": 169.94415194679718,
            "psys": 230.82286463756225,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 8232952,
          "npu_utilization": 0.9900910388365787,
          "seconds": 8.315348465,
          "watts": {
            "core": 2.544483985133869,
            "dram": 2.0051965434980725,
            "package-0": 23.685828420661384,
            "psys": 35.768717841700216,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 807.7971145138014,
            "sha256": "12bab974e4665b4bad98d081d32a834c49bfe99679cebc03d4d2530f6e3343b7",
            "sum": -188106.27967495285,
            "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": 13.443935,
            "dram": 9.319617,
            "package-0": 44.534127,
            "psys": 81.678685,
            "uncore": 0.0
          },
          "net_j": {
            "core": 8.865066262042646,
            "dram": 7.858337017593319,
            "package-0": 34.78666898975571,
            "psys": 57.64308616949319,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2982914,
          "npu_utilization": 0.9940812204035485,
          "seconds": 3.00067433,
          "token_latencies_s": [
            0.216194819,
            0.214429381,
            0.214437496,
            0.214042825,
            0.214269138,
            0.214237467,
            0.214345016,
            0.214175515,
            0.214069859,
            0.214022498,
            0.21397039,
            0.214380772,
            0.214274325,
            0.213793645
          ],
          "watts": {
            "core": 4.480304598733312,
            "dram": 3.1058408794399224,
            "package-0": 14.841373005647034,
            "psys": 27.220109887766462,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 37.574184,
            "dram": 26.835807,
            "package-0": 247.278908,
            "psys": 383.122981,
            "uncore": 0.0
          },
          "net_j": {
            "core": 19.819443516104283,
            "dram": 21.169637396388193,
            "package-0": 209.48276287032724,
            "psys": 289.9240176680551,
            "uncore": 0.0
          }
        },
        "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": 30,
          "gross_j": {
            "core": 3.652518,
            "dram": 0.760923,
            "package-0": 6.108078,
            "psys": 9.606909,
            "uncore": 0.0
          },
          "logits": {
            "l2": 738.8544173581239,
            "sha256": "756a111ecb880182890767495a4326b190317a49fe7bcb1d25e5e6a9434fe988",
            "sum": -54558.6855568951,
            "top1": 78
          },
          "net_j": {
            "core": 3.191353672194068,
            "dram": 0.6137490495356621,
            "package-0": 5.126355181905815,
            "psys": 7.18614496145093,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 214068,
          "npu_utilization": 0.7083295763532781,
          "seconds": 0.30221525,
          "watts": {
            "core": 12.085816318005131,
            "dram": 2.5178180121618614,
            "package-0": 20.21101847110627,
            "psys": 31.788299895521487,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 352.03427535792554,
          "decode_tokens_per_s": 4.5342164871431505,
          "end_to_end_ms": 11635.230677,
          "handoff_estimate_ms": 87.79911499999999,
          "handoff_plus_first_decode_ms": 302.21524999999997,
          "output_tokens_per_s_e2e": 1.3751338881168966,
          "prefill_tokens_per_s": 489.9693413393484,
          "steady_tpot_ms": 214.416135,
          "steady_tpot_p95_ms": 215.8251669,
          "ttft_ms": 8327.051625
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 822,
          "gross_j": {
            "core": 23.016237,
            "dram": 16.746295,
            "package-0": 199.239968,
            "psys": 298.86276,
            "uncore": 0.0
          },
          "logits": {
            "l2": 696.8009887540125,
            "sha256": "ee9a1e49729de1ec15cfac4c7e2d3b98754929839e7fba889ae61a8056468138",
            "sum": 28229.591541170143,
            "top1": 49
          },
          "net_j": {
            "core": 10.309601033587255,
            "dram": 12.691155216776835,
            "package-0": 172.19018613852626,
            "psys": 232.16252855588326,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 8248030,
          "npu_utilization": 0.9905102515801925,
          "seconds": 8.327051625,
          "watts": {
            "core": 2.764031981127534,
            "dram": 2.011071355643241,
            "package-0": 23.926832325841385,
            "psys": 35.89058570295582,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 807.7971145138014,
            "sha256": "12bab974e4665b4bad98d081d32a834c49bfe99679cebc03d4d2530f6e3343b7",
            "sum": -188106.27967495285,
            "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": 10.905429,
            "dram": 9.328589,
            "package-0": 41.930862,
            "psys": 74.653312,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.318488810322959,
            "dram": 7.864733130075695,
            "package-0": 32.16622154989515,
            "psys": 50.57534415072092,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2983659,
          "npu_utilization": 0.992579816834401,
          "seconds": 3.005963802,
          "token_latencies_s": [
            0.217259245,
            0.214424887,
            0.214065245,
            0.214083861,
            0.214407383,
            0.215052971,
            0.21434233,
            0.214384469,
            0.214740951,
            0.215012252,
            0.214780049,
            0.214695391,
            0.214320501,
            0.214336702
          ],
          "watts": {
            "core": 3.6279309127888157,
            "dram": 3.103360391031082,
            "package-0": 13.949223863607921,
            "psys": 24.835066859530997,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 45.51795,
            "dram": 26.996146,
            "package-0": 255.931535,
            "psys": 391.566258,
            "uncore": 0.0
          },
          "net_j": {
            "core": 27.709570088850494,
            "dram": 21.312858146051827,
            "package-0": 218.02120269992076,
            "psys": 298.0857282312882,
            "uncore": 0.0
          }
        },
        "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": 3.75054,
            "dram": 0.785215,
            "package-0": 5.853562,
            "psys": 9.619909,
            "uncore": 0.0
          },
          "logits": {
            "l2": 738.8544173581239,
            "sha256": "756a111ecb880182890767495a4326b190317a49fe7bcb1d25e5e6a9434fe988",
            "sum": -54558.6855568951,
            "top1": 78
          },
          "net_j": {
            "core": 3.239667423621946,
            "dram": 0.6221773782325748,
            "package-0": 4.7660206722588905,
            "psys": 6.938214240431916,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 241067,
          "npu_utilization": 0.7200530655276486,
          "seconds": 0.334790603,
          "watts": {
            "core": 11.202644179352907,
            "dram": 2.3453913967830213,
            "package-0": 17.48424820633332,
            "psys": 28.734106972530526,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 350.97393651755306,
          "decode_tokens_per_s": 4.495824289639343,
          "end_to_end_ms": 11670.382253,
          "handoff_estimate_ms": 120.38274850000002,
          "handoff_plus_first_decode_ms": 334.79060300000003,
          "output_tokens_per_s_e2e": 1.3709919395216916,
          "prefill_tokens_per_s": 489.5636002805126,
          "steady_tpot_ms": 214.40785449999998,
          "steady_tpot_p95_ms": 215.12930295,
          "ttft_ms": 8333.952928
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 824,
          "gross_j": {
            "core": 31.541179,
            "dram": 16.864764,
            "package-0": 208.53035,
            "psys": 305.749889,
            "uncore": 0.0
          },
          "logits": {
            "l2": 696.8009887540125,
            "sha256": "ee9a1e49729de1ec15cfac4c7e2d3b98754929839e7fba889ae61a8056468138",
            "sum": 28229.591541170143,
            "top1": 49
          },
          "net_j": {
            "core": 18.824012013862138,
            "dram": 12.806263393569331,
            "package-0": 181.45814979057826,
            "psys": 238.99437766474102,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 8235990,
          "npu_utilization": 0.9882453226162498,
          "seconds": 8.333952928,
          "watts": {
            "core": 3.7846600853755143,
            "dram": 2.0236212210100932,
            "package-0": 25.021781596508674,
            "psys": 36.6872589324037,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 807.7971145138014,
            "sha256": "12bab974e4665b4bad98d081d32a834c49bfe99679cebc03d4d2530f6e3343b7",
            "sum": -188106.27967495285,
            "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": 10.226231,
            "dram": 9.346167,
            "package-0": 41.547623,
            "psys": 76.19646,
            "uncore": 0.0
          },
          "net_j": {
            "core": 5.645890651366411,
            "dram": 7.88441737424992,
            "package-0": 31.797032237083613,
            "psys": 52.15313632611527,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 2982719,
          "npu_utilization": 0.993696869026465,
          "seconds": 3.001638722,
          "token_latencies_s": [
            0.215038941,
            0.214048186,
            0.213877246,
            0.213981359,
            0.214170559,
            0.215297118,
            0.214493708,
            0.214116809,
            0.214437614,
            0.214098724,
            0.214443708,
            0.214378095,
            0.214455961,
            0.214737187
          ],
          "watts": {
            "core": 3.406882688795484,
            "dram": 3.1136881768944606,
            "package-0": 13.841646796292896,
            "psys": 25.384953705964353,
            "uncore": 0.0
          }
        }
      }
    ]
  },
  {
    "available_devices": [
      "CPU",
      "GPU",
      "NPU"
    ],
    "compile_s": 32.28103984426707,
    "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": 408778,
      "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]"
    },
    "idle_watts": {
      "core": 1.5259465821328744,
      "dram": 0.48698386485903,
      "package-0": 3.248422500499841,
      "psys": 8.010065800945089,
      "uncore": 0.0023192603038971636
    },
    "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": "npu_npu",
      "label": "NPU -> NPU",
      "mode": "staged",
      "prefill_device": "NPU"
    },
    "prompt_length": 8176,
    "prompt_sha256": "998ace73fce48442f97bcb84c31be293d0d94180b17f70a4d127bba1092ad744",
    "properties": {
      "++NPUW_LLM_GENERATE_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "++NPUW_LLM_PREFILL_CONFIG": {
        "NPUW_DEVICES": "NPU"
      },
      "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": 0.045671031810343266,
    "schema_version": 1,
    "trials": [
      {
        "energy": {
          "gross_j": {
            "core": 63.986957000000004,
            "dram": 49.518855,
            "package-0": 489.540861,
            "psys": 774.901154,
            "uncore": 0.0
          },
          "net_j": {
            "core": 29.524571885076412,
            "dram": 38.520681346982556,
            "package-0": 416.1776197574829,
            "psys": 593.999688863384,
            "uncore": 0.0
          }
        },
        "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": 42,
          "gross_j": {
            "core": 5.017077,
            "dram": 0.910642,
            "package-0": 7.468059,
            "psys": 12.535613,
            "uncore": 0.0
          },
          "logits": {
            "l2": 853.2973260506939,
            "sha256": "903a38a0133d569cf197c1de4a4fa2499328b7104c3bb75a4c1e60690bc64ff9",
            "sum": -24865.39645318687,
            "top1": 78
          },
          "net_j": {
            "core": 4.362365329500047,
            "dram": 0.7017002038705891,
            "package-0": 6.074314194844895,
            "psys": 9.098871751809543,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 252885,
          "npu_utilization": 0.5894029674589407,
          "seconds": 0.429052811,
          "watts": {
            "core": 11.693378696917568,
            "dram": 2.1224473459981597,
            "package-0": 17.405920223652842,
            "psys": 29.216946442520804,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 362.7303902253494,
          "decode_tokens_per_s": 4.196490936579461,
          "end_to_end_ms": 22584.26705,
          "handoff_estimate_ms": 204.59441599999997,
          "handoff_plus_first_decode_ms": 429.05281099999996,
          "output_tokens_per_s_e2e": 0.7084577934088856,
          "prefill_tokens_per_s": 430.09277310586225,
          "steady_tpot_ms": 224.458395,
          "steady_tpot_p95_ms": 226.4199107,
          "ttft_ms": 19009.852086
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 1873,
          "gross_j": {
            "core": 47.952758,
            "dram": 38.917686,
            "package-0": 438.461647,
            "psys": 680.977322,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1035.6859878171401,
            "sha256": "db3bb6aec8041a493e2031745915bf05e60ecc0d225663002c7a7306b7dc4726",
            "sum": -316508.9812684683,
            "top1": 8
          },
          "net_j": {
            "core": 18.944739182516813,
            "dram": 29.66019476076123,
            "package-0": 376.7096157526638,
            "psys": 528.7071559249067,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 18831046,
          "npu_utilization": 0.9905940306536272,
          "seconds": 19.009852086,
          "watts": {
            "core": 2.5225213632943153,
            "dram": 2.047237707265557,
            "package-0": 23.064968891731127,
            "psys": 35.82233669779644,
            "uncore": 0.0
          }
        },
        "repeat": 0,
        "steady_decode": {
          "final_logits": {
            "l2": 729.8049311494284,
            "sha256": "ba8fd8377af1d70434f76c9576db141e6f73b12f7b59811353caa97ab51d741c",
            "sum": -13244.23879000172,
            "top1": 730
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 311,
          "gross_j": {
            "core": 11.017122,
            "dram": 9.690527,
            "package-0": 43.611155,
            "psys": 81.388219,
            "uncore": 0.0
          },
          "net_j": {
            "core": 6.217467373059551,
            "dram": 8.15878638235074,
            "package-0": 33.39368980997418,
            "psys": 56.19366118666769,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3128724,
          "npu_utilization": 0.9947102584088351,
          "seconds": 3.145362153,
          "token_latencies_s": [
            0.227211196,
            0.225993834,
            0.224869697,
            0.224925683,
            0.225050445,
            0.224959263,
            0.224047093,
            0.223549384,
            0.225517242,
            0.223962014,
            0.223812155,
            0.22401702,
            0.22373254,
            0.223680659
          ],
          "watts": {
            "core": 3.5026561216462886,
            "dram": 3.08089387759604,
            "package-0": 13.865225331335635,
            "psys": 25.875627365317257,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 67.928658,
            "dram": 49.421689,
            "package-0": 494.684279,
            "psys": 783.1447850000001,
            "uncore": 0.0
          },
          "net_j": {
            "core": 33.443595513671106,
            "dram": 38.41627819064884,
            "package-0": 421.27276235624157,
            "psys": 602.1242808096741,
            "uncore": 0.0
          }
        },
        "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": 4.055592,
            "dram": 0.882933,
            "package-0": 6.412703,
            "psys": 10.812045,
            "uncore": 0.0
          },
          "logits": {
            "l2": 853.2973260506939,
            "sha256": "903a38a0133d569cf197c1de4a4fa2499328b7104c3bb75a4c1e60690bc64ff9",
            "sum": -24865.39645318687,
            "top1": 78
          },
          "net_j": {
            "core": 3.44121692372987,
            "dram": 0.6868640467232092,
            "package-0": 5.104826404010212,
            "psys": 7.587040373359749,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 254460,
          "npu_utilization": 0.6320119130594518,
          "seconds": 0.402618993,
          "watts": {
            "core": 10.073027031787346,
            "dram": 2.1929740408446157,
            "package-0": 15.927472651544782,
            "psys": 26.854284541911813,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 362.49185878054215,
          "decode_tokens_per_s": 4.228329211382308,
          "end_to_end_ms": 22599.128233000003,
          "handoff_estimate_ms": 178.09616499999998,
          "handoff_plus_first_decode_ms": 402.618993,
          "output_tokens_per_s_e2e": 0.7079919116807464,
          "prefill_tokens_per_s": 429.14968292690077,
          "steady_tpot_ms": 224.522828,
          "steady_tpot_p95_ms": 225.88628919999996,
          "ttft_ms": 19051.627731
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 1875,
          "gross_j": {
            "core": 51.557363,
            "dram": 38.867943,
            "package-0": 443.287439,
            "psys": 691.30304,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1035.6859878171401,
            "sha256": "db3bb6aec8041a493e2031745915bf05e60ecc0d225663002c7a7306b7dc4726",
            "sum": -316508.9812684683,
            "top1": 8
          },
          "net_j": {
            "core": 22.485596779812663,
            "dram": 29.590107695702144,
            "package-0": 381.39970280747286,
            "psys": 538.6982482595798,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 18869638,
          "npu_utilization": 0.9904475494918539,
          "seconds": 19.051627731,
          "watts": {
            "core": 2.7061920234830144,
            "dram": 2.0401376485409553,
            "package-0": 23.267693724599788,
            "psys": 36.28577304579288,
            "uncore": 0.0
          }
        },
        "repeat": 1,
        "steady_decode": {
          "final_logits": {
            "l2": 729.8049311494284,
            "sha256": "ba8fd8377af1d70434f76c9576db141e6f73b12f7b59811353caa97ab51d741c",
            "sum": -13244.23879000172,
            "top1": 730
          },
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 310,
          "gross_j": {
            "core": 12.315703,
            "dram": 9.670813,
            "package-0": 44.984137,
            "psys": 81.0297,
            "uncore": 0.0
          },
          "net_j": {
            "core": 7.516781810128572,
            "dram": 8.139306448223483,
            "package-0": 34.768233144758504,
            "psys": 55.83899217673452,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3128244,
          "npu_utilization": 0.9947096547350395,
          "seconds": 3.144881509,
          "token_latencies_s": [
            0.225905459,
            0.224533631,
            0.224512025,
            0.225521625,
            0.225875967,
            0.22478843,
            0.225705447,
            0.223566774,
            0.223798248,
            0.223645236,
            0.223619253,
            0.224820536,
            0.224287232,
            0.224251434
          ],
          "watts": {
            "core": 3.91611034144053,
            "dram": 3.0750961434712676,
            "package-0": 14.30392110839938,
            "psys": 25.76558123672061,
            "uncore": 0.0
          }
        }
      },
      {
        "energy": {
          "gross_j": {
            "core": 59.053499,
            "dram": 49.319086999999996,
            "package-0": 485.008462,
            "psys": 765.983879,
            "uncore": 0.0
          },
          "net_j": {
            "core": 24.515326342677973,
            "dram": 38.29672684482219,
            "package-0": 411.4838848621984,
            "psys": 584.6845865681095,
            "uncore": 0.0
          }
        },
        "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": 39,
          "gross_j": {
            "core": 3.902517,
            "dram": 0.884458,
            "package-0": 6.2769,
            "psys": 11.154146,
            "uncore": 0.0
          },
          "logits": {
            "l2": 853.2973260506939,
            "sha256": "903a38a0133d569cf197c1de4a4fa2499328b7104c3bb75a4c1e60690bc64ff9",
            "sum": -24865.39645318687,
            "top1": 78
          },
          "net_j": {
            "core": 3.2891431155937667,
            "dram": 0.6887085626807496,
            "package-0": 4.971154732993861,
            "psys": 7.93439687365254,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 251861,
          "npu_utilization": 0.6265777560689739,
          "seconds": 0.401962881,
          "watts": {
            "core": 9.708650187528137,
            "dram": 2.2003474494949695,
            "package-0": 15.615620985660117,
            "psys": 27.749194085411087,
            "uncore": 0.0
          }
        },
        "metrics": {
          "all_tokens_per_s_e2e": 361.9344464126539,
          "decode_tokens_per_s": 4.22430750449732,
          "end_to_end_ms": 22633.932971,
          "handoff_estimate_ms": 176.9532425,
          "handoff_plus_first_decode_ms": 401.962881,
          "output_tokens_per_s_e2e": 0.7069032156497147,
          "prefill_tokens_per_s": 428.442927681114,
          "steady_tpot_ms": 225.0096385,
          "steady_tpot_p95_ms": 225.54362394999998,
          "ttft_ms": 19083.055109
        },
        "prefill": {
          "gpu_active_fraction": 0.0,
          "gpu_freq_mhz_after": 0,
          "gpu_freq_mhz_before": 0,
          "gpu_freq_mhz_max": 0,
          "gpu_sample_count": 1875,
          "gross_j": {
            "core": 46.56586,
            "dram": 38.749229,
            "package-0": 437.514444,
            "psys": 676.893287,
            "uncore": 0.0
          },
          "logits": {
            "l2": 1035.6859878171401,
            "sha256": "db3bb6aec8041a493e2031745915bf05e60ecc0d225663002c7a7306b7dc4726",
            "sum": -316508.9812684683,
            "top1": 8
          },
          "net_j": {
            "core": 17.446137279768163,
            "dram": 29.45608906970132,
            "package-0": 375.524618405646,
            "psys": 524.0367598938486,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 18857980,
          "npu_utilization": 0.9882054991868756,
          "seconds": 19.083055109,
          "watts": {
            "core": 2.440167978031908,
            "dram": 2.0305568882272413,
            "package-0": 22.926855343705334,
            "psys": 35.47090773116103,
            "uncore": 0.0
          }
        },
        "repeat": 2,
        "steady_decode": {
          "final_logits": {
            "l2": 729.8049311494284,
            "sha256": "ba8fd8377af1d70434f76c9576db141e6f73b12f7b59811353caa97ab51d741c",
            "sum": -13244.23879000172,
            "top1": 730
          },
          "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": 8.585122,
            "dram": 9.6854,
            "package-0": 41.217118,
            "psys": 77.936446,
            "uncore": 0.0
          },
          "net_j": {
            "core": 3.7800459473160455,
            "dram": 8.15192921244012,
            "package-0": 30.98811172355857,
            "psys": 52.71342980060825,
            "uncore": 0.0
          },
          "npu_busy_delta_us": 3125178,
          "npu_utilization": 0.9924618539581969,
          "seconds": 3.148914981,
          "token_latencies_s": [
            0.225773722,
            0.225231111,
            0.224771474,
            0.224611573,
            0.224468478,
            0.225158248,
            0.225276914,
            0.225019186,
            0.224511305,
            0.224095895,
            0.225376798,
            0.225419725,
            0.225000091,
            0.224143085
          ],
          "watts": {
            "core": 2.7263746566042966,
            "dram": 3.0757896159280267,
            "package-0": 13.089307983447268,
            "psys": 24.750254125708327,
            "uncore": 0.0
          }
        }
      }
    ]
  }
]
