{"device": "GPU", "precision": "fp16", "flop_per_rep": 81077993472, "bytes_per_rep": 215490560, "one_s": 0.006616696355243524, "reps_list": [75, 150, 225], "windows": [{"reps": 75, "dt": 0.5004456536844373, "joules": {"package-0": 15.281394, "core": 2.558832, "uncore": 9.953404, "dram": 1.173215, "psys": 24.709898}}, {"reps": 150, "dt": 0.9925044532865286, "joules": {"package-0": 30.280196, "core": 5.204333, "uncore": 19.622142, "dram": 2.341547, "psys": 50.297845}}, {"reps": 225, "dt": 1.4860328985378146, "joules": {"package-0": 45.271918, "core": 7.874796, "uncore": 29.206529, "dram": 3.504141, "psys": 74.702629}}], "slope_j": {"psys": 0.3332848733333332, "package-0": 0.19993682666666682, "core": 0.035439760000000035, "uncore": 0.12835416666666677, "dram": 0.015539506666666663}, "intercept_j": {"psys": -0.08927366666663754, "package-0": 0.2873119999999763, "core": -0.10331033333333828, "uncore": 0.34089999999998355, "dram": 0.008708333333333762}, "r2": {"psys": 0.999813330611845, "package-0": 0.9999999814229331, "core": 0.9999926502892336, "uncore": 0.9999936018564927, "dram": 0.9999979800436557}, "net_j_per_op": {"psys": 0.27537842952048924, "package-0": 0.16867850389659247, "core": 0.01671614461231036, "uncore": 0.12834160770814124, "dram": 0.012364527051562112}, "pJ_per_flop": {"psys": 3.396463303147607, "package-0": 2.0804474392283154, "core": 0.20617363474939995, "uncore": 1.5829401075701701, "dram": 0.15250164097650196}, "pJ_per_byte": {"psys": 1277.914120787886, "package-0": 782.7651656601221, "core": 77.5725146025439, "uncore": 595.578793373321, "dram": 57.378509070476746}, "watts": {"psys": 50.26983525970645, "package-0": 30.46495003209243, "core": 5.29920704161289, "uncore": 19.65402584878022, "dram": 2.358050756109039}, "steady_cv": 0.014542656924238175, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 57, "summed_node_us": 6059.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "prefill", "seq": 512, "compile_s": 0.4973515560850501, "compile_path": "plain", "flops": 81077993472, "bytes": 215490560, "intensity": 376.2484698726478, "tflops": 12.2535460476055, "gbps": 32.56769669190434, "j_igpu": 0.12834160770814124, "j_npu": 0.011256224524578768, "j_cpu": 0.01671614461231036, "j_dram": 0.012364527051562112, "j_package": 0.16867850389659247, "j_psys": 0.27537842952048924, "watts_psys": 41.618719484121485, "gflops_per_watt": 294.42390826754087}
{"device": "GPU", "precision": "fp16", "flop_per_rep": 11816402944, "bytes_per_rep": 42995712, "one_s": 0.0020901159652405316, "reps_list": [150, 300, 450], "windows": [{"reps": 150, "dt": 0.3137684157118201, "joules": {"package-0": 7.563701, "core": 1.926875, "uncore": 4.122915, "dram": 0.580138, "psys": 12.817289}}, {"reps": 300, "dt": 0.6263203890994191, "joules": {"package-0": 15.121787, "core": 3.829824, "uncore": 8.269266, "dram": 1.160092, "psys": 25.673396}}, {"reps": 450, "dt": 0.9405521843582392, "joules": {"package-0": 22.669315, "core": 5.725266, "uncore": 12.399687, "dram": 1.740902, "psys": 38.602501}}], "slope_j": {"psys": 0.08595070666666671, "package-0": 0.05035204666666665, "core": 0.012661303333333342, "uncore": 0.027589239999999998, "dram": 0.003869213333333331}, "intercept_j": {"psys": -0.0874833333333432, "package-0": 0.012653666666672336, "core": 0.028930666666663996, "uncore": -0.012815999999998459, "dram": -0.0003866666666660912}, "r2": {"psys": 0.9999973284820587, "package-0": 0.9999998371584512, "core": 0.999998697997929, "uncore": 0.9999987652251451, "dram": 0.9999998187248241}, "net_j_per_op": {"psys": 0.06765892233722329, "package-0": 0.04047800742401662, "core": 0.006746791717907528, "uncore": 0.027585272811797356, "dram": 0.0028662844959559926}, "pJ_per_flop": {"psys": 5.72584759151078, "package-0": 3.4255777850373734, "core": 0.5709683183521884, "uncore": 2.3344898563910514, "dram": 0.24256827644925583}, "pJ_per_byte": {"psys": 1573.620233041455, "package-0": 941.4428914217449, "core": 156.91778096168122, "uncore": 641.581951516406, "dram": 66.6644268143761}, "watts": {"psys": 41.04238089281499, "package-0": 24.102134232421996, "core": 6.087132745225065, "uncore": 13.183412049019479, "dram": 1.8509361085454905}, "steady_cv": 0.009070219908780609, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 44, "summed_node_us": 1634.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "prefill", "seq": 512, "compile_s": 0.04175765439867973, "compile_path": "plain", "flops": 11816402944, "bytes": 42995712, "intensity": 274.8274745165285, "tflops": 5.653467625965032, "gbps": 20.570969608881022, "j_igpu": 0.027585272811797356, "j_npu": 0.003279658398355741, "j_cpu": 0.006746791717907528, "j_dram": 0.0028662844959559926, "j_package": 0.04047800742401662, "j_psys": 0.06765892233722329, "watts_psys": 32.37089398981605, "gflops_per_watt": 174.64663248854436}
{"device": "GPU", "precision": "fp16", "flop_per_rep": 69261590528, "bytes_per_rep": 172494848, "one_s": 0.004807903700859556, "reps_list": [91, 182, 273], "windows": [{"reps": 91, "dt": 0.4378730710595846, "joules": {"package-0": 14.461572, "core": 2.286799, "uncore": 9.733678, "dram": 1.089719, "psys": 23.280763}}, {"reps": 182, "dt": 0.8747183354571462, "joules": {"package-0": 28.779528, "core": 4.515796, "uncore": 19.340222, "dram": 2.173457, "psys": 46.760927}}, {"reps": 273, "dt": 1.3125577103346586, "joules": {"package-0": 43.152111, "core": 6.91258, "uncore": 28.822131, "dram": 3.258963, "psys": 70.718264}}], "slope_j": {"psys": 0.26064560989011015, "package-0": 0.15764032417582416, "core": 0.025416379120879122, "uncore": 0.10488160989010993, "dram": 0.011918923076923063}, "intercept_j": {"psys": -0.5175163333333804, "package-0": 0.10719800000000153, "core": -0.054055999999999806, "uncore": 0.2102239999999919, "dram": 0.004802333333335795}, "r2": {"psys": 0.999966273426801, "package-0": 0.9999987915866704, "core": 0.9995616360752335, "uncore": 0.9999857894415984, "dram": 0.9999997785751492}, "net_j_per_op": {"psys": 0.21856893135388591, "package-0": 0.134927024719396, "core": 0.011811199907145818, "uncore": 0.10487248414820916, "dram": 0.00961188109977116}, "pJ_per_flop": {"psys": 3.1557018787422484, "package-0": 1.9480786347932595, "core": 0.1705303013850219, "uncore": 1.5141506764245174, "dram": 0.13877649973812567}, "pJ_per_byte": {"psys": 1267.1041128943511, "package-0": 782.2090125230638, "core": 68.47276915276807, "uncore": 607.9745880190529, "dram": 55.7227141054738}, "watts": {"psys": 53.87821308212741, "package-0": 32.8763532911613, "core": 5.2664960523812105, "uncore": 21.95875333561624, "dram": 2.4829102555567424}, "steady_cv": 0.009376859001790356, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 15, "summed_node_us": 4464.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "prefill", "seq": 512, "compile_s": 0.18887883517891169, "compile_path": "plain", "flops": 69261590528, "bytes": 172494848, "intensity": 401.528459145632, "tflops": 14.405777410977976, "gbps": 35.877350864818986, "j_igpu": 0.10487248414820916, "j_npu": 0.008631459564269854, "j_cpu": 0.011811199907145818, "j_dram": 0.00961188109977116, "j_package": 0.134927024719396, "j_psys": 0.21856893135388591, "watts_psys": 45.46033884056584, "gflops_per_watt": 316.8867144061672}
{"device": "GPU", "precision": "fp16", "flop_per_rep": 158363648, "bytes_per_rep": 154792960, "one_s": 0.002636037036484363, "reps_list": [153, 306, 459], "windows": [{"reps": 153, "dt": 0.40584603045135736, "joules": {"package-0": 8.250955, "core": 2.468804, "uncore": 3.358146, "dram": 1.119015, "psys": 15.071067}}, {"reps": 306, "dt": 0.8063144301995635, "joules": {"package-0": 16.805194, "core": 5.128344, "uncore": 6.851545, "dram": 2.23321, "psys": 30.701704}}, {"reps": 459, "dt": 1.2099409997463226, "joules": {"package-0": 25.19299, "core": 7.674358, "uncore": 10.247532, "dram": 3.347831, "psys": 45.895696}}], "slope_j": {"psys": 0.10073408169934651, "package-0": 0.05536612745098039, "core": 0.01701161437908496, "uncore": 0.02251433333333336, "dram": 0.007283712418300656}, "intercept_j": {"psys": -0.2684733333333706, "package-0": -0.19232199999999958, "core": -0.11505199999999742, "uncore": -0.0703116666666747, "dram": 0.004535999999999281}, "r2": {"psys": 0.9999331177362462, "package-0": 0.9999678289901123, "core": 0.9998414863528932, "uncore": 0.9999333632683173, "dram": 0.9999999878227421}, "net_j_per_op": {"psys": 0.0776646339951622, "package-0": 0.042913070502855435, "core": 0.009552281054876223, "uncore": 0.022509329948254834, "dram": 0.006018826830968844}, "pJ_per_flop": {"psys": 490.41958161485525, "package-0": 270.97803722515556, "core": 60.31864746432352, "uncore": 142.13697545193475, "dram": 38.0063664040427}, "pJ_per_byte": {"psys": 501.7323397340692, "package-0": 277.22882554126124, "core": 61.71004840837867, "uncore": 145.41572141429967, "dram": 38.88307860363187}, "watts": {"psys": 37.93217686616334, "package-0": 20.821668168350346, "core": 6.342753904206083, "uncore": 8.469447685588394, "dram": 2.766937396701086}, "steady_cv": 0.01453506090498214, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 58, "summed_node_us": 2303.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "decode", "seq": 512, "compile_s": 0.28788063302636147, "compile_path": "plain", "flops": 158363648, "bytes": 154792960, "intensity": 1.0230675090133299, "tflops": 0.060076412359974594, "gbps": 58.72184565602488, "j_igpu": 0.022509329948254834, "j_npu": 0.0048326326687555335, "j_cpu": 0.009552281054876223, "j_dram": 0.006018826830968844, "j_package": 0.042913070502855435, "j_psys": 0.0776646339951622, "watts_psys": 29.462649014500258, "gflops_per_watt": 2.039070293048244}
{"device": "GPU", "precision": "fp16", "flop_per_rep": 23087104, "bytes_per_rep": 19449856, "one_s": 0.0007203015395336681, "reps_list": [534, 1068, 1602], "windows": [{"reps": 534, "dt": 0.389630694873631, "joules": {"package-0": 7.442791, "core": 2.818902, "uncore": 2.866448, "dram": 0.704588, "psys": 13.187649}}, {"reps": 1068, "dt": 0.7691654935479164, "joules": {"package-0": 14.853539, "core": 5.561021, "uncore": 5.786423, "dram": 1.400448, "psys": 26.256036}}, {"reps": 1602, "dt": 1.1539230663329363, "joules": {"package-0": 22.334537, "core": 8.366433, "uncore": 8.701699, "dram": 2.104121, "psys": 39.357749}}], "slope_j": {"psys": 0.024503838951310867, "package-0": 0.01394358239700374, "core": 0.005194317415730336, "uncore": 0.005463718164794011, "dram": 0.0013104241573033708}, "intercept_j": {"psys": 0.09704466666666178, "package-0": -0.014790333333327984, "core": 0.03458766666666809, "uncore": -0.05039433333333795, "dram": 0.003519333333333504}, "r2": {"psys": 0.9999994594521725, "package-0": 0.9999925821688957, "core": 0.9999566117845257, "uncore": 0.9999997838424257, "dram": 0.9999896117212225}, "net_j_per_op": {"psys": 0.01820007369022897, "package-0": 0.01054076370838701, "core": 0.0031560420793230056, "uncore": 0.005462350981433133, "dram": 0.0009647920499851183}, "pJ_per_flop": {"psys": 788.3220732331334, "package-0": 456.56500305915415, "core": 136.7015143745619, "uncore": 236.5974953564177, "dram": 41.789219210218754}, "pJ_per_byte": {"psys": 935.7433643842386, "package-0": 541.9455911852001, "core": 162.26557560750092, "uncore": 280.8427466729385, "dram": 49.604071618068446}, "watts": {"psys": 34.10777559467234, "package-0": 19.35530855707491, "core": 7.250425304858298, "uncore": 7.540969804558302, "dram": 1.8234499867367306}, "steady_cv": 0.009403549450796975, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 45, "summed_node_us": 462.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "decode", "seq": 512, "compile_s": 0.04238830413669348, "compile_path": "plain", "flops": 23087104, "bytes": 19449856, "intensity": 1.187006423080973, "tflops": 0.032051998687864625, "gbps": 27.002380159553834, "j_igpu": 0.005462350981433133, "j_npu": 0.000957578597645753, "j_cpu": 0.0031560420793230056, "j_dram": 0.0009647920499851183, "j_package": 0.01054076370838701, "j_psys": 0.01820007369022897, "watts_psys": 25.267298056883117, "gflops_per_watt": 1.268517061686114}
{"device": "GPU", "precision": "fp16", "flop_per_rep": 135276544, "bytes_per_rep": 135343104, "one_s": 0.0019023882767627226, "reps_list": [194, 388, 582], "windows": [{"reps": 194, "dt": 0.371729482896626, "joules": {"package-0": 7.655498, "core": 2.051447, "uncore": 3.175834, "dram": 1.162045, "psys": 13.920923}}, {"reps": 388, "dt": 0.7381266513839364, "joules": {"package-0": 15.436423, "core": 4.074513, "uncore": 6.513716, "dram": 2.320062, "psys": 28.068105}}, {"reps": 582, "dt": 1.1049995962530375, "joules": {"package-0": 23.17914, "core": 6.129074, "uncore": 9.770483, "dram": 3.47869, "psys": 42.070387}}], "slope_j": {"psys": 0.07255016494845354, "package-0": 0.040009386597938165, "core": 0.010509347938144321, "uncore": 0.016996518041237106, "dram": 0.005970734536082472}, "intercept_j": {"psys": -0.12965899999997058, "package-0": -0.09995500000000608, "core": 0.007384333333336694, "uncore": -0.10797133333333046, "dram": 0.0036206666666673826}, "r2": {"psys": 0.9999911677486284, "package-0": 0.9999979807074583, "core": 0.9999801144163685, "uncore": 0.9999495715268968, "dram": 0.9999999768131272}, "net_j_per_op": {"psys": 0.055901291561468676, "package-0": 0.031022202723022977, "core": 0.005126059284747441, "uncore": 0.01699290717348913, "dram": 0.005057885634689412}, "pJ_per_flop": {"psys": 413.2371356372667, "package-0": 229.32432930148613, "core": 37.89318630691394, "uncore": 125.61606521740481, "dram": 37.389228650677325}, "pJ_per_byte": {"psys": 413.0339109221898, "package-0": 229.21155054211687, "core": 37.87455092464439, "uncore": 125.5542888501296, "dram": 37.37084110830953}, "watts": {"psys": 38.072762327386556, "package-0": 20.976604949538945, "core": 5.546675329822006, "uncore": 8.842069294080199, "dram": 3.14813689687847}, "steady_cv": 0.008581159721183, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 15, "summed_node_us": 1818.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "decode", "seq": 512, "compile_s": 0.1879706410691142, "compile_path": "plain", "flops": 135276544, "bytes": 135343104, "intensity": 0.9995082128454805, "tflops": 0.07110879816301166, "gbps": 71.14378576297376, "j_igpu": 0.01699290717348913, "j_npu": 0.0038453506300969945, "j_cpu": 0.005126059284747441, "j_dram": 0.005057885634689412, "j_package": 0.031022202723022977, "j_psys": 0.055901291561468676, "watts_psys": 29.38479607149147, "gflops_per_watt": 2.4199180416296975}
{"device": "GPU", "precision": "int8", "flop_per_rep": 81077993472, "bytes_per_rep": 138416128, "one_s": 0.0061622465463976065, "reps_list": [75, 150, 225], "windows": [{"reps": 75, "dt": 0.46424277406185865, "joules": {"package-0": 11.334505, "core": 2.394769, "uncore": 6.618513, "dram": 0.930418, "psys": 20.511117}}, {"reps": 150, "dt": 0.924336981959641, "joules": {"package-0": 22.509525, "core": 4.847033, "uncore": 13.029386, "dram": 1.855403, "psys": 40.844988}}, {"reps": 225, "dt": 1.3792393123731017, "joules": {"package-0": 33.674719, "core": 7.211651, "uncore": 19.538524, "dram": 2.778802, "psys": 60.9631}}], "slope_j": {"psys": 0.2696798866666666, "package-0": 0.14893476000000008, "core": 0.032112546666666644, "uncore": 0.0861334066666667, "dram": 0.012322559999999996}, "intercept_j": {"psys": 0.3210853333333432, "package-0": 0.16603566666665395, "core": 0.0009356666666698507, "uncore": 0.14212999999999454, "dram": 0.0064903333333340045}, "r2": {"psys": 0.9999905172819902, "package-0": 0.9999999355151671, "core": 0.9998896523711851, "uncore": 0.9999807184331932, "dram": 0.9999997545852826}, "net_j_per_op": {"psys": 0.2157505895912333, "package-0": 0.11982333018753397, "core": 0.014674911922035645, "uncore": 0.0861217102860796, "dram": 0.009365645232019725}, "pJ_per_flop": {"psys": 2.6610252714966607, "package-0": 1.4778773506391045, "core": 0.18099747284820972, "uncore": 1.062208209627455, "dram": 0.11551402336139606}, "pJ_per_byte": {"psys": 1558.7099040310773, "package-0": 865.6746285196907, "core": 106.02024586351415, "uncore": 622.1941874149203, "dram": 67.66296216593868}, "watts": {"psys": 44.200523761976925, "package-0": 24.41542863367178, "core": 5.228716246197858, "uncore": 14.166159436379653, "dram": 2.0147352059004384}, "steady_cv": 0.009547152677837768, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": true, "int8_verified": true, "n_nodes": 90, "summed_node_us": 5512.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "prefill", "seq": 512, "compile_s": 0.5296273315325379, "compile_path": "plain", "flops": 81077993472, "bytes": 138416128, "intensity": 585.7553931287545, "tflops": 13.157213503474226, "gbps": 22.461958793407383, "j_igpu": 0.0861217102860796, "j_npu": 0.009661062747399005, "j_cpu": 0.014674911922035645, "j_dram": 0.009365645232019725, "j_package": 0.11982333018753397, "j_psys": 0.2157505895912333, "watts_psys": 35.011677635222036, "gflops_per_watt": 375.79500304315496}
{"device": "GPU", "precision": "int8", "flop_per_rep": 11816402944, "bytes_per_rep": 33556480, "one_s": 0.00244928906840677, "reps_list": [158, 316, 474], "windows": [{"reps": 158, "dt": 0.39039830584079027, "joules": {"package-0": 7.84904, "core": 2.139338, "uncore": 3.919973, "dram": 0.649595, "psys": 14.428613}}, {"reps": 316, "dt": 0.7729118252173066, "joules": {"package-0": 16.115621, "core": 4.779895, "uncore": 7.78599, "dram": 1.297604, "psys": 29.393784}}, {"reps": 474, "dt": 1.160963018424809, "joules": {"package-0": 23.835632, "core": 6.857038, "uncore": 11.640717, "dram": 1.963923, "psys": 43.588573}}], "slope_j": {"psys": 0.0922783544303799, "package-0": 0.050590481012658194, "core": 0.014929430379746849, "uncore": 0.024432734177215202, "dram": 0.004159265822784811}, "intercept_j": {"psys": -0.02297000000005293, "package-0": -0.05316099999998869, "core": -0.12560966666667106, "uncore": 0.06148266666666302, "dram": -0.010620666666666908}, "r2": {"psys": 0.9997673965664701, "package-0": 0.9996105162343973, "core": 0.9952683396541945, "uncore": 0.9999992872322943, "dram": 0.9999353125815745}, "net_j_per_op": {"psys": 0.0708432435825451, "package-0": 0.0390196510777464, "core": 0.00799854765096954, "uncore": 0.024428085253012603, "dram": 0.002983990058841477}, "pJ_per_flop": {"psys": 5.995330721056452, "package-0": 3.3021598250048982, "core": 0.6769020732346431, "uncore": 2.067302999803034, "dram": 0.2525294772853573}, "pJ_per_byte": {"psys": 2111.1643289923463, "package-0": 1162.8052488743276, "core": 238.3607473420794, "uncore": 727.9692403080597, "dram": 88.92440622024351}, "watts": {"psys": 37.54518646006557, "package-0": 20.530914096075268, "core": 5.906336283909894, "uncore": 10.026776749352523, "dram": 1.6916326952986365}, "steady_cv": 0.011267902763037045, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": true, "int8_verified": true, "n_nodes": 61, "summed_node_us": 1962.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "prefill", "seq": 512, "compile_s": 0.3216154668480158, "compile_path": "plain", "flops": 11816402944, "bytes": 33556480, "intensity": 352.134757400061, "tflops": 4.824421541915595, "gbps": 13.700498006888195, "j_igpu": 0.024428085253012603, "j_npu": 0.003609028114922782, "j_cpu": 0.00799854765096954, "j_dram": 0.002983990058841477, "j_package": 0.0390196510777464, "j_psys": 0.0708432435825451, "watts_psys": 28.9240026815731, "gflops_per_watt": 166.7964698741069}
{"device": "GPU", "precision": "int8", "flop_per_rep": 69261590528, "bytes_per_rep": 104859648, "one_s": 0.003930116978850454, "reps_list": [94, 188, 282], "windows": [{"reps": 94, "dt": 0.3714521387591958, "joules": {"package-0": 9.716467, "core": 1.893733, "uncore": 5.902145, "dram": 0.806761, "psys": 17.215532}}, {"reps": 188, "dt": 0.7388619920238853, "joules": {"package-0": 19.530467, "core": 3.936819, "uncore": 11.773468, "dram": 1.61114, "psys": 34.640537}}, {"reps": 282, "dt": 1.107566167600453, "joules": {"package-0": 29.338182, "core": 5.89592, "uncore": 17.723526, "dram": 2.413691, "psys": 51.782582}}], "slope_j": {"psys": 0.18386728723404241, "package-0": 0.10437082446808509, "core": 0.021288228723404242, "uncore": 0.06287968617021286, "dram": 0.008547499999999996}, "intercept_j": {"psys": -0.020832999999977158, "package-0": -0.09334299999999729, "core": -0.09336299999999713, "uncore": -0.02166800000001956, "dram": 0.0036006666666672515}, "r2": {"psys": 0.9999776645604818, "package-0": 0.9999999658008456, "core": 0.9998532345939676, "uncore": 0.9999852132899286, "dram": 0.999999568642828}, "net_j_per_op": {"psys": 0.14947261566454254, "package-0": 0.085804328674198, "core": 0.010166968990469661, "uncore": 0.06287222652970832, "dram": 0.006661658392210633}, "pJ_per_flop": {"psys": 2.158088119621164, "package-0": 1.2388443294485183, "core": 0.1467908679682936, "uncore": 0.9077502559559517, "dram": 0.09618113504796805}, "pJ_per_byte": {"psys": 1425.454104752884, "package-0": 818.2778629411192, "core": 96.9578783105362, "uncore": 599.5845659305314, "dram": 63.52928432689983}, "watts": {"psys": 46.753488427862685, "package-0": 26.48887520965118, "core": 5.323311755516636, "uncore": 16.002227693897602, "dram": 2.1792747653436115}, "steady_cv": 0.010372732642872442, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": true, "int8_verified": true, "n_nodes": 33, "summed_node_us": 3570.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "prefill", "seq": 512, "compile_s": 0.37763929180800915, "compile_path": "plain", "flops": 69261590528, "bytes": 104859648, "intensity": 660.5170992754048, "tflops": 17.623289815729297, "gbps": 26.68105009705617, "j_igpu": 0.06287222652970832, "j_npu": 0.006103474761809384, "j_cpu": 0.010166968990469661, "j_dram": 0.006661658392210633, "j_package": 0.085804328674198, "j_psys": 0.14947261566454254, "watts_psys": 38.03261237996605, "gflops_per_watt": 463.37310831197317}
{"device": "GPU", "precision": "int8", "flop_per_rep": 158363648, "bytes_per_rep": 77718528, "one_s": 0.0018951223860612981, "reps_list": [197, 394, 591], "windows": [{"reps": 197, "dt": 0.37849793396890163, "joules": {"package-0": 7.437176, "core": 2.353266, "uncore": 3.100823, "dram": 0.862547, "psys": 13.574306}}, {"reps": 394, "dt": 0.7466782201081514, "joules": {"package-0": 15.052818, "core": 4.803149, "uncore": 6.306442, "dram": 1.722957, "psys": 27.390494}}, {"reps": 591, "dt": 1.1150102503597736, "joules": {"package-0": 22.344181, "core": 6.952986, "uncore": 9.511084, "dram": 2.578058, "psys": 40.517413}}], "slope_j": {"psys": 0.06838352030456854, "package-0": 0.03783503807106599, "core": 0.011674416243654832, "uncore": 0.016269697969543136, "dram": 0.004354088832487312}, "intercept_j": {"psys": 0.21763066666666, "package-0": 0.037720000000002564, "core": 0.10341366666666296, "uncore": -0.104144666666663, "dram": 0.005676333333332432}, "r2": {"psys": 0.9997818947325783, "package-0": 0.9998422871419765, "core": 0.9985836320846411, "uncore": 0.9999999922568719, "dram": 0.9999968076110767}, "net_j_per_op": {"psys": 0.05179823482748057, "package-0": 0.028882179415952856, "core": 0.0063116882651988015, "uncore": 0.016266100892970554, "dram": 0.00344472642248271}, "pJ_per_flop": {"psys": 327.0841224084492, "package-0": 182.3788462864461, "core": 39.85566349923185, "uncore": 102.7136031431314, "dram": 21.75200221759674}, "pJ_per_byte": {"psys": 666.4850217888913, "package-0": 371.62540463906953, "core": 81.21214371428653, "uncore": 209.29502026814706, "dram": 44.323104298664916}, "watts": {"psys": 36.33815293350576, "package-0": 20.0394399897134, "core": 6.2358045567352605, "uncore": 8.530041761437722, "dram": 2.3121383854257425}, "steady_cv": 0.007751100765815719, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": true, "int8_verified": true, "n_nodes": 86, "summed_node_us": 1507.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "decode", "seq": 512, "compile_s": 0.5018099024891853, "compile_path": "plain", "flops": 158363648, "bytes": 77718528, "intensity": 2.0376562973503565, "tflops": 0.08356381053000643, "gbps": 41.00976726971457, "j_igpu": 0.016266100892970554, "j_npu": 0.0028596638353007896, "j_cpu": 0.0063116882651988015, "j_dram": 0.00344472642248271, "j_package": 0.028882179415952856, "j_psys": 0.05179823482748057, "watts_psys": 27.33239563231308, "gflops_per_watt": 3.05731746511144}
{"device": "GPU", "precision": "int8", "flop_per_rep": 23087104, "bytes_per_rep": 10010624, "one_s": 0.0009532108224560449, "reps_list": [216, 432, 648], "windows": [{"reps": 216, "dt": 0.21815333422273397, "joules": {"package-0": 3.900991, "core": 1.259641, "uncore": 1.799861, "dram": 0.306579, "psys": 6.974408}}, {"reps": 432, "dt": 0.41148151736706495, "joules": {"package-0": 7.932535, "core": 2.652093, "uncore": 3.646169, "dram": 0.594114, "psys": 13.884608}}, {"reps": 648, "dt": 0.6176806129515171, "joules": {"package-0": 11.845795, "core": 3.96111, "uncore": 5.437914, "dram": 0.890501, "psys": 20.745857}}], "slope_j": {"psys": 0.03187835416666665, "package-0": 0.018390750000000004, "core": 0.006253400462962966, "uncore": 0.008421418981481483, "dram": 0.0013516712962962966}, "intercept_j": {"psys": 0.09684200000000705, "package-0": -0.05169700000000077, "core": -0.07718766666666799, "uncore": -0.010071666666667293, "dram": 0.013142666666666525}, "r2": {"psys": 0.9999957884614872, "package-0": 0.9999261190811388, "core": 0.9996821390024027, "uncore": 0.9999250270461308, "dram": 0.9999234017915568}, "net_j_per_op": {"psys": 0.023536268332758944, "package-0": 0.013887630927194607, "core": 0.0035560494187508937, "uncore": 0.008419609719798224, "dram": 0.0008942791525875048}, "pJ_per_flop": {"psys": 1019.4552046354079, "package-0": 601.5319603183929, "core": 154.02752197724297, "uncore": 364.68886352303974, "dram": 38.735007759635195}, "pJ_per_byte": {"psys": 2351.1289938328464, "package-0": 1387.2892366344604, "core": 355.2275481279582, "uncore": 841.067421950742, "dram": 89.33300787118813}, "watts": {"psys": 33.586705758609234, "package-0": 19.177864338976427, "core": 6.412877330036769, "uncore": 8.803763443400857, "dram": 1.4416852032069478}, "steady_cv": 0.008685094583737143, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": true, "int8_verified": true, "n_nodes": 57, "summed_node_us": 654.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "decode", "seq": 512, "compile_s": 0.34511128440499306, "compile_path": "plain", "flops": 23087104, "bytes": 10010624, "intensity": 2.3062602291325693, "tflops": 0.02422035446525221, "gbps": 10.502004136090909, "j_igpu": 0.008419609719798224, "j_npu": 0.0010176926360579847, "j_cpu": 0.0035560494187508937, "j_dram": 0.0008942791525875048, "j_package": 0.013887630927194607, "j_psys": 0.023536268332758944, "watts_psys": 24.69156641771581, "gflops_per_watt": 0.98091607699196}
{"device": "GPU", "precision": "int8", "flop_per_rep": 135276544, "bytes_per_rep": 67707904, "one_s": 0.0009620255423766193, "reps_list": [462, 924, 1386], "windows": [{"reps": 462, "dt": 0.44708281196653843, "joules": {"package-0": 8.914955, "core": 2.57055, "uncore": 3.444877, "dram": 1.39233, "psys": 16.728168}}, {"reps": 924, "dt": 0.8889116011559963, "joules": {"package-0": 18.070266, "core": 5.175829, "uncore": 7.122358, "dram": 2.782524, "psys": 33.672033}}, {"reps": 1386, "dt": 1.3332682643085718, "joules": {"package-0": 27.197379, "core": 7.827678, "uncore": 10.690951, "dram": 4.18163, "psys": 50.203912}}], "slope_j": {"psys": 0.03622916017316017, "package-0": 0.019786173160173164, "core": 0.005689532467532468, "uncore": 0.007842071428571422, "dram": 0.003018722943722946}, "intercept_j": {"psys": 0.058960333333336244, "package-0": -0.22155733333333671, "core": -0.06577566666666712, "uncore": -0.16001199999999352, "dram": -0.0038053333333352888}, "r2": {"psys": 0.9999495151279609, "package-0": 0.9999992070469758, "core": 0.9999738433116033, "uncore": 0.9999247336997827, "dram": 0.9999965971941602}, "net_j_per_op": {"psys": 0.027809931750806163, "package-0": 0.01524141195274166, "core": 0.0029672379444780806, "uncore": 0.007840245435925163, "dram": 0.0025571011127868567}, "pJ_per_flop": {"psys": 205.57837248419182, "package-0": 112.66854919609463, "core": 21.93460785395346, "uncore": 57.95716836116957, "dram": 18.902767894387193}, "pJ_per_byte": {"psys": 410.73390413630534, "package-0": 225.10535775471146, "core": 43.82409983446069, "uncore": 115.79512837858876, "dram": 37.766655910465886}, "watts": {"psys": 37.65477161945017, "package-0": 20.399029758729363, "core": 5.871045017380209, "uncore": 8.018604572084591, "dram": 3.1363755606742645}, "steady_cv": 0.00937618440047832, "gpu_mhz": 2450.0, "npu_util": 0.0, "int8_kernel_seen": true, "int8_verified": true, "n_nodes": 33, "summed_node_us": 855.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "decode", "seq": 512, "compile_s": 0.3459257259964943, "compile_path": "plain", "flops": 135276544, "bytes": 67707904, "intensity": 1.99794316480392, "tflops": 0.14061637455675907, "gbps": 70.38056789296071, "j_igpu": 0.007840245435925163, "j_npu": 0.0018768274595515597, "j_cpu": 0.0029672379444780806, "j_dram": 0.0025571011127868567, "j_package": 0.01524141195274166, "j_psys": 0.027809931750806163, "watts_psys": 28.907685426006047, "gflops_per_watt": 4.864324918599578}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 81077993472, "bytes_per_rep": 215490560, "one_s": 0.013975925217180149, "reps_list": [23, 46, 69], "windows": [{"reps": 23, "dt": 0.35138389468193054, "joules": {"package-0": 5.998581, "core": 0.63745, "uncore": 0.0, "dram": 0.589781, "psys": 9.79685}}, {"reps": 46, "dt": 0.6428925599902868, "joules": {"package-0": 13.18063, "core": 1.167965, "uncore": 0.0, "dram": 1.148922, "psys": 20.702706}}, {"reps": 69, "dt": 0.962806642986834, "joules": {"package-0": 19.817576, "core": 1.744868, "uncore": 0.0, "dram": 1.724544, "psys": 30.850079}}], "slope_j": {"psys": 0.457678891304348, "package-0": 0.3004129347826087, "core": 0.0240743043478261, "uncore": 0.0, "dram": 0.02466876086956524}, "intercept_j": {"psys": -0.6033506666666758, "package-0": -0.8200660000000018, "core": 0.07600966666666602, "uncore": 0.0, "dram": 0.019652666666665652}, "r2": {"psys": 0.9995675407768231, "package-0": 0.999481608612751, "core": 0.9994154617301868, "uncore": 1.0, "dram": 0.9999296919593719}, "net_j_per_op": {"psys": 0.33536768267833794, "package-0": 0.2343884479398904, "core": 0.0, "uncore": 0.0, "dram": 0.017962502143999245}, "pJ_per_flop": {"psys": 4.136358934365537, "package-0": 2.8909009449134384, "core": 0.0, "uncore": 0.0, "dram": 0.22154596302636093}, "pJ_per_byte": {"psys": 1556.2987198990895, "package-0": 1087.6970570770727, "core": 0.0, "uncore": 0.0, "dram": 83.3563295951305}, "watts": {"psys": 32.04182192210099, "package-0": 20.583131768307705, "core": 1.8122724980241545, "uncore": 0.0, "dram": 1.791163379024985}, "steady_cv": 0.014206326353461163, "gpu_mhz": 0.0, "npu_util": 0.9804736842105264, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 12, "summed_node_us": 27995.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "prefill", "seq": 512, "compile_s": 3.3576742904260755, "compile_path": "plain", "flops": 81077993472, "bytes": 215490560, "intensity": 376.2484698726478, "tflops": 5.801261255486218, "gbps": 15.41869727058245, "j_igpu": 0.0, "j_npu": 0.21642594579589114, "j_cpu": 0.0, "j_dram": 0.017962502143999245, "j_package": 0.2343884479398904, "j_psys": 0.33536768267833794, "watts_psys": 23.996098824719052, "gflops_per_watt": 241.7585165764602}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 11816402944, "bytes_per_rep": 42995712, "one_s": 0.0022649606871709693, "reps_list": [128, 256, 384], "windows": [{"reps": 128, "dt": 0.34351586271077394, "joules": {"package-0": 4.965015, "core": 0.851805, "uncore": 0.0, "dram": 0.408019, "psys": 8.285441}}, {"reps": 256, "dt": 0.5798299359157681, "joules": {"package-0": 11.236238, "core": 1.499752, "uncore": 0.0, "dram": 0.747679, "psys": 17.59346}}, {"reps": 384, "dt": 0.8680355418473482, "joules": {"package-0": 16.751789, "core": 2.242548, "uncore": 0.0, "dram": 1.120847, "psys": 26.206049}}], "slope_j": {"psys": 0.07000237499999999, "package-0": 0.046042085937499956, "core": 0.005432589843750002, "uncore": 0.0, "dram": 0.0027844843750000022}, "intercept_j": {"psys": -0.558957999999997, "package-0": -0.8024266666666572, "core": 0.1406253333333328, "uncore": 0.0, "dram": 0.04602033333333274}, "r2": {"psys": 0.9994982791386021, "package-0": 0.9986317644523625, "core": 0.9984519760567018, "uncore": 1.0, "dram": 0.9992639852921673}, "net_j_per_op": {"psys": 0.050180425838613846, "package-0": 0.03534205250389473, "core": 0.0, "uncore": 0.0, "dram": 0.0016976574111055179}, "pJ_per_flop": {"psys": 4.2466752425782754, "package-0": 2.990931561100861, "core": 0.0, "uncore": 0.0, "dram": 0.14366955994569694}, "pJ_per_byte": {"psys": 1167.1030320096536, "package-0": 821.9901673891278, "core": 0.0, "uncore": 0.0, "dram": 39.4843423247769}, "watts": {"psys": 30.190064503843285, "package-0": 19.298505870334456, "core": 2.583474860058636, "uncore": 0.0, "dram": 1.2912455146878201}, "steady_cv": 0.011625544620011783, "gpu_mhz": 0.0, "npu_util": 0.8406470588235294, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 6, "summed_node_us": 1867.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "prefill", "seq": 512, "compile_s": 0.8273244798183441, "compile_path": "plain", "flops": 11816402944, "bytes": 42995712, "intensity": 274.8274745165285, "tflops": 5.217045492634656, "gbps": 18.982983785781926, "j_igpu": 0.0, "j_npu": 0.033644395092789214, "j_cpu": 0.0, "j_dram": 0.0016976574111055179, "j_package": 0.03534205250389473, "j_psys": 0.050180425838613846, "watts_psys": 22.15509793297618, "gflops_per_watt": 235.47833137173728}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 69261590528, "bytes_per_rep": 172494848, "one_s": 0.8878968660719693, "reps_list": [2, 4, 6], "windows": [{"reps": 2, "dt": 1.7753295237198472, "joules": {"package-0": 24.68762, "core": 4.49065, "uncore": 0.0, "dram": 6.053024, "psys": 48.799253}}, {"reps": 4, "dt": 3.554432215169072, "joules": {"package-0": 49.031003, "core": 8.583597, "uncore": 0.0, "dram": 12.124419, "psys": 95.404908}}, {"reps": 6, "dt": 5.327381196431816, "joules": {"package-0": 75.599782, "core": 14.790672, "uncore": 0.0, "dram": 18.185501, "psys": 147.786304}}], "slope_j": {"psys": 24.746762750000016, "package-0": 12.728040499999983, "core": 2.575005500000001, "uncore": 0.0, "dram": 3.0331192499999986}, "intercept_j": {"psys": -1.6568960000000554, "package-0": -1.1393603333332578, "core": -1.0117156666666698, "uncore": 0.0, "dram": -0.011495666666661464}, "r2": {"psys": 0.9988664392108921, "package-0": 0.9993635361068736, "core": 0.9861513144300137, "uncore": 1.0, "dram": 0.9999997591482962}, "net_j_per_op": {"psys": 16.976276207111887, "package-0": 8.53347491377344, "core": 0.06247681019665885, "uncore": 0.0, "dram": 2.607067592383928}, "pJ_per_flop": {"psys": 245.10375920762291, "package-0": 123.20645322639044, "core": 0.9020412283399946, "uncore": 0.0, "dram": 37.64088541007419}, "pJ_per_byte": {"psys": 98416.13476543882, "package-0": 49470.89732079094, "core": 362.1952245011912, "uncore": 0.0, "dram": 15113.886719584389}, "watts": {"psys": 27.74089154704841, "package-0": 14.190796418066608, "core": 2.7763494772828583, "uncore": 0.0, "dram": 3.4135910927831334}, "steady_cv": 0.025486733851912073, "gpu_mhz": 0.0, "npu_util": 0.9993386792452831, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 7, "summed_node_us": 9476.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["overhead_bound:0.01"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "prefill", "seq": 512, "compile_s": 1.3415083279833198, "compile_path": "plain", "flops": 69261590528, "bytes": 172494848, "intensity": 401.528459145632, "tflops": 0.07800634642896233, "gbps": 0.19427351823316186, "j_igpu": 0.0, "j_npu": 5.863930511192853, "j_cpu": 0.06247681019665885, "j_dram": 2.607067592383928, "j_package": 8.53347491377344, "j_psys": 16.976276207111887, "watts_psys": 19.119648751790795, "gflops_per_watt": 4.079904784948314}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 158363648, "bytes_per_rep": 154792960, "one_s": 0.002775328526466759, "reps_list": [122, 244, 366], "windows": [{"reps": 122, "dt": 0.3798884302377701, "joules": {"package-0": 4.15862, "core": 1.188656, "uncore": 0.0, "dram": 0.927366, "psys": 7.706706}}, {"reps": 244, "dt": 0.6769684422761202, "joules": {"package-0": 8.585305, "core": 2.044977, "uncore": 0.0, "dram": 1.797786, "psys": 15.724386}}, {"reps": 366, "dt": 1.0157702406868339, "joules": {"package-0": 12.857573, "core": 3.048393, "uncore": 0.0, "dram": 2.698113, "psys": 22.821048}}], "slope_j": {"psys": 0.06194402459016396, "package-0": 0.035651446721311514, "core": 0.007621872950819669, "uncore": 0.0, "dram": 0.007257159836065568}, "intercept_j": {"psys": 0.3030379999999961, "package-0": -0.1651203333333425, "core": 0.23427166666666763, "uncore": 0.0, "dram": 0.03700800000000163}, "r2": {"psys": 0.998763770020918, "package-0": 0.9998949757994796, "core": 0.9979190252729406, "uncore": 1.0, "dram": 0.9999049241702123}, "net_j_per_op": {"psys": 0.03765555845061182, "package-0": 0.022540354687513645, "core": 0.0, "uncore": 0.0, "dram": 0.005925436114186268}, "pJ_per_flop": {"psys": 237.77905426005225, "package-0": 142.3328836647767, "core": 0.0, "uncore": 0.0, "dram": 37.41664320707154}, "pJ_per_byte": {"psys": 243.26402473737707, "package-0": 145.61614874160713, "core": 0.0, "uncore": 0.0, "dram": 38.2797519614992}, "watts": {"psys": 22.466742070105422, "package-0": 12.657954018524986, "core": 3.001065475140093, "uncore": 0.0, "dram": 2.656223712732139}, "steady_cv": 0.011098873956317607, "gpu_mhz": 0.0, "npu_util": 0.96893, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 16, "summed_node_us": 4897.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "decode", "seq": 512, "compile_s": 1.1756198965013027, "compile_path": "plain", "flops": 158363648, "bytes": 154792960, "intensity": 1.0230675090133299, "tflops": 0.05706122590164526, "gbps": 55.774643803004196, "j_igpu": 0.0, "j_npu": 0.016614918573327378, "j_cpu": 0.0, "j_dram": 0.005925436114186268, "j_package": 0.022540354687513645, "j_psys": 0.03765555845061182, "watts_psys": 13.567964329812407, "gflops_per_watt": 4.205584899443364}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 23087104, "bytes_per_rep": 19449856, "one_s": 0.000724520388308082, "reps_list": [453, 906, 1359], "windows": [{"reps": 453, "dt": 0.3927381942048669, "joules": {"package-0": 3.727286, "core": 1.412533, "uncore": 0.0, "dram": 0.576415, "psys": 6.943219}}, {"reps": 906, "dt": 0.6564154718071222, "joules": {"package-0": 7.255231, "core": 2.290644, "uncore": 0.0, "dram": 1.079221, "psys": 13.165371}}, {"reps": 1359, "dt": 0.9720462299883366, "joules": {"package-0": 10.811129, "core": 3.369803, "uncore": 0.0, "dram": 1.613277, "psys": 19.512707}}], "slope_j": {"psys": 0.01387360706401766, "package-0": 0.007818811258278153, "core": 0.0021603421633554073, "uncore": 0.0, "dram": 0.0011444392935982353}, "intercept_j": {"psys": 0.6376109999999997, "package-0": 0.1807056666666599, "core": 0.40039000000000097, "uncore": 0.0, "dram": 0.052775666666665444}, "r2": {"psys": 0.999966938172169, "package-0": 0.9999948096638513, "core": 0.9964952904573621, "uncore": 1.0, "dram": 0.9996973048981536}, "net_j_per_op": {"psys": 0.007532920276482928, "package-0": 0.004396062059082039, "core": 0.00011012852759293866, "uncore": 0.0, "dram": 0.0007967827985585547}, "pJ_per_flop": {"psys": 326.2825981328333, "package-0": 190.4120178555976, "core": 4.770131740773492, "uncore": 0.0, "dram": 34.51202881741056}, "pJ_per_byte": {"psys": 387.29953972322096, "package-0": 226.02028822640327, "core": 5.662177015240558, "uncore": 0.0, "dram": 40.96599987982197}, "watts": {"psys": 20.0738466937258, "package-0": 11.122031716670225, "core": 3.4667106316954017, "uncore": 0.0, "dram": 1.6596710631955822}, "steady_cv": 0.011326984625438095, "gpu_mhz": 0.0, "npu_util": 0.867878947368421, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 10, "summed_node_us": 818.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "decode", "seq": 512, "compile_s": 0.1970375720411539, "compile_path": "plain", "flops": 23087104, "bytes": 19449856, "intensity": 1.187006423080973, "tflops": 0.03186536137915122, "gbps": 26.845146546420573, "j_igpu": 0.0, "j_npu": 0.003489150732930545, "j_cpu": 0.00011012852759293866, "j_dram": 0.0007967827985585547, "j_package": 0.004396062059082039, "j_psys": 0.007532920276482928, "watts_psys": 10.397112901231104, "gflops_per_watt": 3.0648278692230124}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 135276544, "bytes_per_rep": 135343104, "one_s": 0.00217817377276896, "reps_list": [168, 336, 504], "windows": [{"reps": 168, "dt": 0.3990514660254121, "joules": {"package-0": 4.482166, "core": 1.185849, "uncore": 0.0, "dram": 1.07269, "psys": 8.791604}}, {"reps": 336, "dt": 0.7318663876503706, "joules": {"package-0": 9.860144, "core": 2.531, "uncore": 0.0, "dram": 2.102777, "psys": 18.207839}}, {"reps": 504, "dt": 1.0877144178375602, "joules": {"package-0": 14.208948, "core": 3.298392, "uncore": 0.0, "dram": 3.149467, "psys": 25.825129}}], "slope_j": {"psys": 0.050695014880952384, "package-0": 0.028948755952380945, "core": 0.006287330357142855, "uncore": 0.0, "dram": 0.006180883928571429}, "intercept_j": {"psys": 0.5746656666666657, "package-0": -0.20969599999999744, "core": 0.22587066666666752, "uncore": 0.0, "dram": 0.03153433333333311}, "r2": {"psys": 0.9962958057429632, "package-0": 0.9962820776905125, "core": 0.9756742678712762, "uncore": 1.0, "dram": 0.9999786959144241}, "net_j_per_op": {"psys": 0.03163258698406063, "package-0": 0.01865871766392803, "core": 0.00012363690964779187, "uncore": 0.0, "dram": 0.005135701112948286}, "pJ_per_flop": {"psys": 233.83645123326502, "package-0": 137.93017704479522, "core": 0.9139567436598016, "uncore": 0.0, "dram": 37.964461251673356}, "pJ_per_byte": {"psys": 233.72145347029004, "package-0": 137.86234475550398, "core": 0.9135072714734831, "uncore": 0.0, "dram": 37.94579081730153}, "watts": {"psys": 23.74256383522236, "package-0": 13.063123708747208, "core": 3.0324062510428025, "uncore": 0.0, "dram": 2.8954907173716835}, "steady_cv": 0.022055815020344025, "gpu_mhz": 0.0, "npu_util": 0.9724809523809524, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 6, "summed_node_us": 2071.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "decode", "seq": 512, "compile_s": 1.0528603754937649, "compile_path": "plain", "flops": 135276544, "bytes": 135343104, "intensity": 0.9995082128454805, "tflops": 0.062105487491951744, "gbps": 62.136045200814145, "j_igpu": 0.0, "j_npu": 0.013399379641331952, "j_cpu": 0.00012363690964779187, "j_dram": 0.005135701112948286, "j_package": 0.01865871766392803, "j_psys": 0.03163258698406063, "watts_psys": 14.522526797229924, "gflops_per_watt": 4.276493227321705}
{"device": "NPU", "precision": "int8", "flop_per_rep": 81077993472, "bytes_per_rep": 138416128, "one_s": 0.008602266338326642, "reps_list": [37, 74, 111], "windows": [{"reps": 37, "dt": 0.3482053242623806, "joules": {"package-0": 4.894946, "core": 0.700498, "uncore": 0.0, "dram": 0.561766, "psys": 9.285072}}, {"reps": 74, "dt": 0.6365677090361714, "joules": {"package-0": 10.622043, "core": 1.202512, "uncore": 0.0, "dram": 1.088682, "psys": 19.072155}}, {"reps": 111, "dt": 0.9547521369531751, "joules": {"package-0": 16.079366, "core": 1.948116, "uncore": 0.0, "dram": 1.63409, "psys": 28.548755}}], "slope_j": {"psys": 0.2603200405405408, "package-0": 0.15114081081081093, "core": 0.016859702702702685, "uncore": 0.0, "dram": 0.014490864864864879}, "intercept_j": {"psys": -0.2950223333333544, "package-0": -0.652301666666677, "core": 0.03609066666666797, "uncore": 0.0, "dram": 0.022521999999999043}, "r2": {"psys": 0.9999134159000316, "package-0": 0.9998061044788844, "core": 0.9874526882450533, "uncore": 1.0, "dram": 0.9999008822547535}, "net_j_per_op": {"psys": 0.18503675328667116, "package-0": 0.11050234059531269, "core": 0.0, "uncore": 0.0, "dram": 0.010363122136238193}, "pJ_per_flop": {"psys": 2.2822068648081792, "package-0": 1.3629141011423067, "core": 0.0, "uncore": 0.0, "dram": 0.127816707005916}, "pJ_per_byte": {"psys": 1336.8149792968572, "package-0": 798.334285115335, "core": 0.0, "uncore": 0.0, "dram": 74.86932546067314}, "watts": {"psys": 29.901745065588837, "package-0": 16.841403519988766, "core": 2.040441623117879, "uncore": 0.0, "dram": 1.7115332207736578}, "steady_cv": 0.01178559852055313, "gpu_mhz": 0.0, "npu_util": 0.9612473684210526, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 0, "summed_node_us": null, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "prefill", "seq": 512, "compile_s": 3.601064680144191, "compile_path": "plain", "flops": 81077993472, "bytes": 138416128, "intensity": 585.7553931287545, "tflops": 9.425189860811926, "gbps": 16.09065827028618, "j_igpu": 0.0, "j_npu": 0.1001392184590745, "j_cpu": 0.0, "j_dram": 0.010363122136238193, "j_package": 0.11050234059531269, "j_psys": 0.18503675328667116, "watts_psys": 21.510233002465426, "gflops_per_watt": 438.1723740385167}
{"device": "NPU", "precision": "int8", "flop_per_rep": 11816402944, "bytes_per_rep": 33556480, "one_s": 0.002002252008145054, "reps_list": [150, 300, 450], "windows": [{"reps": 150, "dt": 0.3742391671985388, "joules": {"package-0": 4.225514, "core": 0.960874, "uncore": 0.0, "dram": 0.449035, "psys": 8.494302}}, {"reps": 300, "dt": 0.6006756024435163, "joules": {"package-0": 8.917824, "core": 1.608517, "uncore": 0.0, "dram": 0.814512, "psys": 16.30428}}, {"reps": 450, "dt": 0.900135450065136, "joules": {"package-0": 13.135891, "core": 2.23559, "uncore": 0.0, "dram": 1.219235, "psys": 24.145873}}], "slope_j": {"psys": 0.05217190333333333, "package-0": 0.029701256666666637, "core": 0.004249053333333338, "uncore": 0.0, "dram": 0.002567333333333331}, "intercept_j": {"psys": 0.6632473333333332, "package-0": -0.1506339999999907, "core": 0.3269443333333317, "uncore": 0.0, "dram": 0.05739400000000069}, "r2": {"psys": 0.9999986399694759, "package-0": 0.9990566376175354, "core": 0.9999132073878225, "uncore": 1.0, "dram": 0.9991352575072657}, "net_j_per_op": {"psys": 0.034649066076803625, "package-0": 0.020242300689635992, "core": 0.0, "uncore": 0.0, "dram": 0.001606565456305865}, "pJ_per_flop": {"psys": 2.932285420615022, "package-0": 1.7130679095463988, "core": 0.0, "uncore": 0.0, "dram": 0.13596061880418767}, "pJ_per_byte": {"psys": 1032.5596152160067, "package-0": 603.2307527379508, "core": 0.0, "uncore": 0.0, "dram": 47.8764595185748}, "watts": {"psys": 26.824710656882527, "package-0": 14.593238161045047, "core": 2.483615104635894, "uncore": 0.0, "dram": 1.3545017029512318}, "steady_cv": 0.008942421473913265, "gpu_mhz": 0.0, "npu_util": 0.8885882352941177, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 0, "summed_node_us": null, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "prefill", "seq": 512, "compile_s": 0.7276036273688078, "compile_path": "plain", "flops": 11816402944, "bytes": 33556480, "intensity": 352.134757400061, "tflops": 5.901556295576933, "gbps": 16.759368882385452, "j_igpu": 0.0, "j_npu": 0.018635735233330127, "j_cpu": 0.0, "j_dram": 0.001606565456305865, "j_package": 0.020242300689635992, "j_psys": 0.034649066076803625, "watts_psys": 17.30504748445904, "gflops_per_watt": 341.0309218091936}
{"device": "NPU", "precision": "int8", "flop_per_rep": 69261590528, "bytes_per_rep": 104859648, "one_s": 0.45301505783572793, "reps_list": [2, 4, 6], "windows": [{"reps": 2, "dt": 0.9063212005421519, "joules": {"package-0": 12.577238, "core": 2.405267, "uncore": 0.0, "dram": 3.063408, "psys": 26.48504}}, {"reps": 4, "dt": 1.8111096899956465, "joules": {"package-0": 24.814267, "core": 4.506031, "uncore": 0.0, "dram": 6.116683, "psys": 51.549795}}, {"reps": 6, "dt": 2.7180903470143676, "joules": {"package-0": 37.276272, "core": 6.802473, "uncore": 0.0, "dram": 9.172523, "psys": 75.456838}}], "slope_j": {"psys": 12.24294949999999, "package-0": 6.174758499999996, "core": 1.0993015000000004, "uncore": 0.0, "dram": 1.5272787500000007}, "intercept_j": {"psys": 2.1920930000000474, "package-0": 0.19022500000001705, "core": 0.17405099999999804, "uncore": 0.0, "dram": 0.00842299999999696}, "r2": {"psys": 0.9998137454806846, "package-0": 0.9999723446486484, "core": 0.9993403363698532, "uncore": 1.0, "dram": 0.9999999412379609}, "net_j_per_op": {"psys": 8.278359078354237, "package-0": 4.034643533667691, "core": 0.0, "uncore": 0.0, "dram": 1.3099023590123704}, "pJ_per_flop": {"psys": 119.52308653679547, "package-0": 58.25225067617569, "core": 0.0, "uncore": 0.0, "dram": 18.91239212132767}, "pJ_per_byte": {"psys": 78947.04241572732, "package-0": 38476.6076428913, "core": 0.0, "uncore": 0.0, "dram": 12491.958384338373}, "watts": {"psys": 27.760974937011962, "package-0": 13.71414016496743, "core": 2.5026662588578197, "uncore": 0.0, "dram": 3.3746203506720724}, "steady_cv": 0.014101292784667607, "gpu_mhz": 0.0, "npu_util": 0.9979716981132075, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 17, "summed_node_us": 2839.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core", "overhead_bound:0.01"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "prefill", "seq": 512, "compile_s": 1.7389692151919007, "compile_path": "plain", "flops": 69261590528, "bytes": 104859648, "intensity": 660.5170992754048, "tflops": 0.15289026121757657, "gbps": 0.23147055751516354, "j_igpu": 0.0, "j_npu": 2.724741174655321, "j_cpu": 0.0, "j_dram": 1.3099023590123704, "j_package": 4.034643533667691, "j_psys": 8.278359078354237, "watts_psys": 18.27391592214167, "gflops_per_watt": 8.366584473135637}
{"device": "NPU", "precision": "int8", "flop_per_rep": 158363648, "bytes_per_rep": 77718528, "one_s": 0.0015551140574618764, "reps_list": [218, 436, 654], "windows": [{"reps": 218, "dt": 0.37469623517245054, "joules": {"package-0": 3.934133, "core": 1.160398, "uncore": 0.0, "dram": 0.864499, "psys": 7.799113}}, {"reps": 436, "dt": 0.6780297290533781, "joules": {"package-0": 8.292032, "core": 2.091303, "uncore": 0.0, "dram": 1.687251, "psys": 15.661887}}, {"reps": 654, "dt": 0.9978044796735048, "joules": {"package-0": 12.297759, "core": 3.064019, "uncore": 0.0, "dram": 2.522088, "psys": 23.025575}}], "slope_j": {"psys": 0.034923077981651375, "package-0": 0.01918262844036696, "core": 0.004366103211009177, "uncore": 0.0, "dram": 0.003801809633027522}, "intercept_j": {"psys": 0.26906300000000033, "package-0": -0.18898466666666044, "core": 0.2016189999999988, "uncore": 0.0, "dram": 0.033690333333333676}, "r2": {"psys": 0.9996420057826254, "package-0": 0.9994093331664783, "core": 0.9998392211012828, "uncore": 1.0, "dram": 0.9999822821652065}, "net_j_per_op": {"psys": 0.021313397265440598, "package-0": 0.011836023043566308, "core": 0.0, "uncore": 0.0, "dram": 0.003055598054542255}, "pJ_per_flop": {"psys": 134.58516228068072, "package-0": 74.73952004165949, "core": 0.0, "uncore": 0.0, "dram": 19.294819822174436}, "pJ_per_byte": {"psys": 274.23830345114874, "package-0": 152.29345367383064, "core": 0.0, "uncore": 0.0, "dram": 39.31621111689422}, "watts": {"psys": 23.076239352557607, "package-0": 12.324818389294055, "core": 3.0707609180133053, "uncore": 0.0, "dram": 2.527637479464175}, "steady_cv": 0.012154514232183916, "gpu_mhz": 0.0, "npu_util": 0.9404684210526316, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 0, "summed_node_us": null, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "decode", "seq": 512, "compile_s": 1.947987417690456, "compile_path": "plain", "flops": 158363648, "bytes": 77718528, "intensity": 2.0376562973503565, "tflops": 0.10183410486203663, "gbps": 49.97609508259832, "j_igpu": 0.0, "j_npu": 0.008780424989024052, "j_cpu": 0.0, "j_dram": 0.003055598054542255, "j_package": 0.011836023043566308, "j_psys": 0.021313397265440598, "watts_psys": 13.705359528565058, "gflops_per_watt": 7.43023958253641}
{"device": "NPU", "precision": "int8", "flop_per_rep": 23087104, "bytes_per_rep": 10010624, "one_s": 0.0006107013867376677, "reps_list": [459, 918, 1377], "windows": [{"reps": 459, "dt": 0.32833366841077805, "joules": {"package-0": 2.763726, "core": 1.123776, "uncore": 0.0, "dram": 0.402465, "psys": 5.600877}}, {"reps": 918, "dt": 0.5606238730251789, "joules": {"package-0": 5.608933, "core": 1.950679, "uncore": 0.0, "dram": 0.746031, "psys": 10.827181}}, {"reps": 1377, "dt": 0.8343506399542093, "joules": {"package-0": 8.382486, "core": 2.91668, "uncore": 0.0, "dram": 1.11517, "psys": 16.201802}}], "slope_j": {"psys": 0.011547848583878, "package-0": 0.006120653594771248, "core": 0.0019530544662309385, "uncore": 0.0, "dram": 0.0007763671023965137}, "intercept_j": {"psys": 0.2756949999999965, "package-0": -0.03371166666667236, "core": 0.20414099999999843, "uncore": 0.0, "dram": 0.04185033333333384}, "r2": {"psys": 0.9999347553734032, "package-0": 0.9999457930318293, "core": 0.9979976685399937, "uncore": 1.0, "dram": 0.9995710208858186}, "net_j_per_op": {"psys": 0.006203256110698217, "package-0": 0.003235603406900653, "core": 0.0002249204748708386, "uncore": 0.0, "dram": 0.0004833259305326112}, "pJ_per_flop": {"psys": 268.6892262753361, "package-0": 140.1476515590978, "core": 9.74225588756557, "uncore": 0.0, "dram": 20.934887742204964}, "pJ_per_byte": {"psys": 619.6672765552095, "package-0": 323.2169549970764, "core": 22.468177295525095, "uncore": 0.0, "dram": 48.28129900120224}, "watts": {"psys": 19.41845697018844, "package-0": 10.046718488116754, "core": 3.4957485022844503, "uncore": 0.0, "dram": 1.3365723553123932}, "steady_cv": 0.0138627818265333, "gpu_mhz": 0.0, "npu_util": 0.751625, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 21, "summed_node_us": 1061.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "decode", "seq": 512, "compile_s": 0.41515316627919674, "compile_path": "plain", "flops": 23087104, "bytes": 10010624, "intensity": 2.3062602291325693, "tflops": 0.03780424361459208, "gbps": 16.39201124706166, "j_igpu": 0.0, "j_npu": 0.002527357001497203, "j_cpu": 0.0002249204748708386, "j_dram": 0.0004833259305326112, "j_package": 0.003235603406900653, "j_psys": 0.006203256110698217, "watts_psys": 10.15759296672906, "gflops_per_watt": 3.7217718546528618}
{"device": "NPU", "precision": "int8", "flop_per_rep": 135276544, "bytes_per_rep": 67707904, "one_s": 0.0010930673730968822, "reps_list": [327, 654, 981], "windows": [{"reps": 327, "dt": 0.40117333084344864, "joules": {"package-0": 4.091237, "core": 0.916685, "uncore": 0.0, "dram": 1.05084, "psys": 9.08628}}, {"reps": 654, "dt": 0.7123537743464112, "joules": {"package-0": 8.642556, "core": 1.657466, "uncore": 0.0, "dram": 2.047358, "psys": 17.98085}}, {"reps": 981, "dt": 1.0722990930080414, "joules": {"package-0": 12.949185, "core": 2.454766, "uncore": 0.0, "dram": 3.075554, "psys": 26.835381}}], "slope_j": {"psys": 0.027139298165137604, "package-0": 0.013544262996941909, "core": 0.002351805810397553, "uncore": 0.0, "dram": 0.0030958929663608536}, "intercept_j": {"psys": 0.21840266666667438, "package-0": -0.2969553333333425, "core": 0.13822466666666688, "uncore": 0.0, "dram": 0.033203333333335174}, "r2": {"psys": 0.9999983037426932, "package-0": 0.999745706955049, "core": 0.9995501022336999, "uncore": 1.0, "dram": 0.9999184110305697}, "net_j_per_op": {"psys": 0.017573248731158706, "package-0": 0.0083804394023133, "core": 0.0, "uncore": 0.0, "dram": 0.0025713915473455705}, "pJ_per_flop": {"psys": 129.90610353823578, "package-0": 61.95042506639807, "core": 0.0, "uncore": 0.0, "dram": 19.00840656711019}, "pJ_per_byte": {"psys": 259.5450116305285, "package-0": 123.77342831810745, "core": 0.0, "uncore": 0.0, "dram": 37.97771597457175}, "watts": {"psys": 25.026022287047443, "package-0": 12.076094332668518, "core": 2.2892549438924044, "uncore": 0.0, "dram": 2.868186702809172}, "steady_cv": 0.009753039515463392, "gpu_mhz": 0.0, "npu_util": 0.9543952380952382, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 16, "summed_node_us": 2325.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "decode", "seq": 512, "compile_s": 1.7120932387188077, "compile_path": "plain", "flops": 135276544, "bytes": 67707904, "intensity": 1.99794316480392, "tflops": 0.12375865141481081, "gbps": 61.943029008513655, "j_igpu": 0.0, "j_npu": 0.00580904785496773, "j_cpu": 0.0, "j_dram": 0.0025713915473455705, "j_package": 0.0083804394023133, "j_psys": 0.017573248731158706, "watts_psys": 16.077004184444842, "gflops_per_watt": 7.697867711855941}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 81077993472, "bytes_per_rep": 215490560, "one_s": 0.08674977137707174, "reps_list": [6, 12, 18], "windows": [{"reps": 6, "dt": 0.5184201458469033, "joules": {"package-0": 7.664531, "core": 1.455257, "uncore": 0.0, "dram": 1.583492, "psys": 15.089683}}, {"reps": 12, "dt": 1.0409972565248609, "joules": {"package-0": 14.852135, "core": 2.497552, "uncore": 0.0, "dram": 3.16613, "psys": 29.542344}}, {"reps": 18, "dt": 1.5622965255752206, "joules": {"package-0": 22.356022, "core": 3.809805, "uncore": 0.0, "dram": 4.753527, "psys": 43.96113}}], "slope_j": {"psys": 2.405953916666666, "package-0": 1.224290916666667, "core": 0.1962123333333331, "uncore": 0.0, "dram": 0.26416958333333307}, "intercept_j": {"psys": 0.659605333333341, "package-0": 0.2660716666666616, "core": 0.232990000000003, "uncore": 0.0, "dram": -0.0023186666666633235}, "r2": {"psys": 0.9999995411190251, "package-0": 0.99984553445938, "core": 0.9956372813874999, "uncore": 1.0, "dram": 0.9999992487547931}, "net_j_per_op": {"psys": 1.6467577117444618, "package-0": 0.8144712410538113, "core": 0.0, "uncore": 0.0, "dram": 0.22254325790600493}, "pJ_per_flop": {"psys": 20.31078522328212, "package-0": 10.045527845173995, "core": 0.0, "uncore": 0.0, "dram": 2.7448047044092116}, "pJ_per_byte": {"psys": 7641.901862171882, "package-0": 3779.614480809792, "core": 0.0, "uncore": 0.0, "dram": 1032.7285701332112}, "watts": {"psys": 28.138787535108925, "package-0": 14.309717543389374, "core": 2.438592762406145, "uncore": 0.0, "dram": 3.0426535053899597}, "steady_cv": 0.012739956466670782, "gpu_mhz": 0.0, "npu_util": 0.97897, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 26, "summed_node_us": 16078.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core", "overhead_bound:0.19"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "prefill", "seq": 512, "compile_s": 3.4388818349689245, "compile_path": "npuw", "flops": 81077993472, "bytes": 215490560, "intensity": 376.2484698726478, "tflops": 0.934619102563182, "gbps": 2.4840475839796263, "j_igpu": 0.0, "j_npu": 0.5919279831478064, "j_cpu": 0.0, "j_dram": 0.22254325790600493, "j_package": 0.8144712410538113, "j_psys": 1.6467577117444618, "watts_psys": 18.98284785773747, "gflops_per_watt": 49.23492563220582}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 11816402944, "bytes_per_rep": 42995712, "one_s": 0.02173936553299427, "reps_list": [16, 32, 48], "windows": [{"reps": 16, "dt": 0.36545317713171244, "joules": {"package-0": 2.945305, "core": 1.275082, "uncore": 0.0, "dram": 0.301818, "psys": 6.196517}}, {"reps": 32, "dt": 0.6956596970558167, "joules": {"package-0": 5.798935, "core": 2.323786, "uncore": 0.0, "dram": 0.579711, "psys": 12.03659}}, {"reps": 48, "dt": 1.037831773981452, "joules": {"package-0": 8.775613, "core": 3.608206, "uncore": 0.0, "dram": 0.8728, "psys": 18.4261}}], "slope_j": {"psys": 0.38217446875000033, "package-0": 0.18219712499999977, "core": 0.072910125, "uncore": 0.0, "dram": 0.017843187500000007}, "intercept_j": {"psys": -0.00984733333334257, "package-0": 0.009643000000007618, "core": 0.0692339999999998, "uncore": 0.0, "dram": 0.013794333333333167}, "r2": {"psys": 0.9993276442036159, "package-0": 0.9998515499351236, "core": 0.9966091692964203, "uncore": 1.0, "dram": 0.9997639576356938}, "net_j_per_op": {"psys": 0.19192101276881257, "package-0": 0.07949691508203172, "core": 0.01139309336386999, "uncore": 0.0, "dram": 0.0074116913715936514}, "pJ_per_flop": {"psys": 16.241915046258985, "package-0": 6.727674695825921, "core": 0.964176104848815, "uncore": 0.0, "dram": 0.6272375279278265}, "pJ_per_byte": {"psys": 4463.724493475363, "package-0": 1848.949846022592, "core": 264.98208388478344, "uncore": 0.0, "dram": 172.3821057223951}, "watts": {"psys": 17.754418839299586, "package-0": 8.455718180928267, "core": 3.476677136370355, "uncore": 0.0, "dram": 0.8409840803501923}, "steady_cv": 0.02432476065345018, "gpu_mhz": 0.0, "npu_util": 0.94482, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 20, "summed_node_us": 2701.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["overhead_bound:0.12"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "prefill", "seq": 512, "compile_s": 1.202588957734406, "compile_path": "npuw", "flops": 11816402944, "bytes": 42995712, "intensity": 274.8274745165285, "tflops": 0.543548657207406, "gbps": 1.9777813632483683, "j_igpu": 0.0, "j_npu": 0.060692130346568084, "j_cpu": 0.01139309336386999, "j_dram": 0.0074116913715936514, "j_package": 0.07949691508203172, "j_psys": 0.19192101276881257, "watts_psys": 8.828271113870835, "gflops_per_watt": 61.56909435567641}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 69261590528, "bytes_per_rep": 172494848, "one_s": 0.02808684946614362, "reps_list": [14, 28, 42], "windows": [{"reps": 14, "dt": 0.40336622204631567, "joules": {"package-0": 5.036913, "core": 1.122189, "uncore": 0.0, "dram": 0.522703, "psys": 9.052833}}, {"reps": 28, "dt": 0.7754938257858157, "joules": {"package-0": 10.178074, "core": 2.109431, "uncore": 0.0, "dram": 1.034055, "psys": 17.988602}}, {"reps": 42, "dt": 1.179647677578032, "joules": {"package-0": 15.428245, "core": 3.258781, "uncore": 0.0, "dram": 1.5545, "psys": 27.150443}}], "slope_j": {"psys": 0.6463432142857143, "package-0": 0.37111900000000025, "core": 0.07630685714285702, "uncore": 0.0, "dram": 0.03684989285714289}, "intercept_j": {"psys": -0.033650666666666496, "package-0": -0.1769213333333397, "core": 0.026875000000003386, "uncore": 0.0, "dram": 0.005288999999999137}, "r2": {"psys": 0.9999479875562916, "package-0": 0.9999633179961398, "core": 0.9980848098878964, "uncore": 1.0, "dram": 0.9999741123062184}, "net_j_per_op": {"psys": 0.4005393443223944, "package-0": 0.23843226948206417, "core": 0.0, "uncore": 0.0, "dram": 0.02337259698019856}, "pJ_per_flop": {"psys": 5.7829937382172965, "package-0": 3.44248908615049, "core": 0.0, "uncore": 0.0, "dram": 0.33745394528226796}, "pJ_per_byte": {"psys": 2322.036564955229, "package-0": 1382.257338387661, "core": 0.0, "uncore": 0.0, "dram": 135.49736268180345}, "watts": {"psys": 23.015721995692257, "package-0": 13.078688911316442, "core": 2.7625036372645564, "uncore": 0.0, "dram": 1.317766337820109}, "steady_cv": 0.03333189432311069, "gpu_mhz": 0.0, "npu_util": 0.9576304347826088, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 16, "summed_node_us": 14302.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "prefill", "seq": 512, "compile_s": 3.0394735308364034, "compile_path": "npuw", "flops": 69261590528, "bytes": 172494848, "intensity": 401.528459145632, "tflops": 2.465979340669345, "gbps": 6.141480845259213, "j_igpu": 0.0, "j_npu": 0.2150596725018656, "j_cpu": 0.0, "j_dram": 0.02337259698019856, "j_package": 0.23843226948206417, "j_psys": 0.4005393443223944, "watts_psys": 14.260743085664041, "gflops_per_watt": 172.92081666826542}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 158363648, "bytes_per_rep": 154792960, "one_s": 0.0035422146320343018, "reps_list": [93, 186, 279], "windows": [{"reps": 93, "dt": 0.3707681791856885, "joules": {"package-0": 3.52587, "core": 1.052488, "uncore": 0.0, "dram": 0.76306, "psys": 7.388775}}, {"reps": 186, "dt": 0.6588519215583801, "joules": {"package-0": 7.243451, "core": 1.909602, "uncore": 0.0, "dram": 1.475094, "psys": 14.426233}}, {"reps": 279, "dt": 0.983666680753231, "joules": {"package-0": 10.931246, "core": 2.939689, "uncore": 0.0, "dram": 2.209528, "psys": 21.74323}}], "slope_j": {"psys": 0.07717448924731175, "package-0": 0.0398138494623656, "core": 0.01014624193548386, "uncore": 0.0, "dram": 0.007776709677419359}, "intercept_j": {"psys": 0.16495766666668268, "package-0": -0.17185366666666843, "core": 0.08005866666666872, "uncore": 0.0, "dram": 0.036092666666665885}, "r2": {"psys": 0.9998736034422563, "package-0": 0.9999946072964703, "core": 0.9972075550196003, "uncore": 1.0, "dram": 0.9999200676822959}, "net_j_per_op": {"psys": 0.04617457002019183, "package-0": 0.023079865863244682, "core": 0.00012264997507932703, "uncore": 0.0, "dram": 0.006077000540831656}, "pJ_per_flop": {"psys": 291.5730384045702, "package-0": 145.73967040241888, "core": 0.7744831381967598, "uncore": 0.0, "dram": 38.37370897664378}, "pJ_per_byte": {"psys": 298.29890209601155, "package-0": 149.1015215630264, "core": 0.7923485349677855, "uncore": 0.0, "dram": 39.258894854337406}, "watts": {"psys": 22.104266033846322, "package-0": 11.112754161429487, "core": 2.988501143241904, "uncore": 0.0, "dram": 2.246216165732157}, "steady_cv": 0.014718443592398872, "gpu_mhz": 0.0, "npu_util": 0.8866842105263159, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 32, "summed_node_us": 2978.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "decode", "seq": 512, "compile_s": 1.8323994828388095, "compile_path": "npuw", "flops": 158363648, "bytes": 154792960, "intensity": 1.0230675090133299, "tflops": 0.044707524656418515, "gbps": 43.69948636091034, "j_igpu": 0.0, "j_npu": 0.0168802153473337, "j_cpu": 0.00012264997507932703, "j_dram": 0.006077000540831656, "j_package": 0.023079865863244682, "j_psys": 0.04617457002019183, "watts_psys": 13.035508803619186, "gflops_per_watt": 3.4296723917677765}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 23087104, "bytes_per_rep": 19449856, "one_s": 0.001394050887310198, "reps_list": [233, 466, 699], "windows": [{"reps": 233, "dt": 0.3582892585545778, "joules": {"package-0": 3.033745, "core": 1.427548, "uncore": 0.0, "dram": 0.406188, "psys": 6.362594}}, {"reps": 466, "dt": 0.6094789085909724, "joules": {"package-0": 5.814194, "core": 2.370477, "uncore": 0.0, "dram": 0.762999, "psys": 12.195342}}, {"reps": 699, "dt": 0.9744415702298284, "joules": {"package-0": 9.311133, "core": 3.977529, "uncore": 0.0, "dram": 1.17053, "psys": 19.080945}}], "slope_j": {"psys": 0.02729259871244636, "package-0": 0.013470789699570824, "core": 0.005472062231759655, "uncore": 0.0, "dram": 0.0016402188841201694}, "intercept_j": {"psys": -0.1720573333333372, "package-0": -0.2243640000000049, "core": 0.041870333333334564, "uncore": 0.0, "dram": 0.015563666666667864}, "r2": {"psys": 0.9977209002672699, "package-0": 0.9956762634888825, "core": 0.9778898426403932, "uncore": 1.0, "dram": 0.9985343684554998}, "net_j_per_op": {"psys": 0.015092472691202907, "package-0": 0.006885072257826905, "core": 0.001527242756376156, "uncore": 0.0, "dram": 0.0009712924422832615}, "pJ_per_flop": {"psys": 653.7187466735935, "package-0": 298.2215637711384, "core": 66.1513352378954, "uncore": 0.0, "dram": 42.070778659950655}, "pJ_per_byte": {"psys": 775.9683511899989, "package-0": 353.9909116975933, "core": 78.52205982276455, "uncore": 0.0, "dram": 49.938284493379356}, "watts": {"psys": 19.581415225850467, "package-0": 9.555352813821262, "core": 4.081854799217847, "uncore": 0.0, "dram": 1.2012315933154647}, "steady_cv": 0.011972400063165736, "gpu_mhz": 0.0, "npu_util": 0.7214578947368422, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 20, "summed_node_us": 753.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "decode", "seq": 512, "compile_s": 0.6090962160378695, "compile_path": "npuw", "flops": 23087104, "bytes": 19449856, "intensity": 1.187006423080973, "tflops": 0.016561163017905504, "gbps": 13.952041619892535, "j_igpu": 0.0, "j_npu": 0.004386537059167487, "j_cpu": 0.001527242756376156, "j_dram": 0.0009712924422832615, "j_package": 0.006885072257826905, "j_psys": 0.015092472691202907, "watts_psys": 10.826342731522251, "gflops_per_watt": 1.5297098409498533}
{"device": "NPU", "precision": "fp16", "flop_per_rep": 135276544, "bytes_per_rep": 135343104, "one_s": 0.002985668765167061, "reps_list": [119, 238, 357], "windows": [{"reps": 119, "dt": 0.37883033882826567, "joules": {"package-0": 3.642081, "core": 1.11633, "uncore": 0.0, "dram": 0.822751, "psys": 8.422952}}, {"reps": 238, "dt": 0.7045734701678157, "joules": {"package-0": 7.672404, "core": 2.115045, "uncore": 0.0, "dram": 1.614071, "psys": 15.019859}}, {"reps": 357, "dt": 1.065883749164641, "joules": {"package-0": 11.469759, "core": 3.092949, "uncore": 0.0, "dram": 2.425226, "psys": 22.733829}}], "slope_j": {"psys": 0.060129735294117664, "package-0": 0.03288940336134451, "core": 0.008305121848739496, "uncore": 0.0, "dram": 0.006733088235294114}, "intercept_j": {"psys": 1.0813363333333281, "package-0": -0.23292999999999253, "core": 0.13148899999999988, "uncore": 0.0, "dram": 0.01820766666666736}, "r2": {"psys": 0.997973155453585, "package-0": 0.9997048266307912, "core": 0.9999630510020098, "uncore": 1.0, "dram": 0.999948933222804}, "net_j_per_op": {"psys": 0.03400046302414261, "package-0": 0.01878463068696259, "core": 0.0, "uncore": 0.0, "dram": 0.005300434089188764}, "pJ_per_flop": {"psys": 251.34041733164483, "package-0": 138.86095942074473, "core": 0.0, "uncore": 0.0, "dram": 39.1822109913509}, "pJ_per_byte": {"psys": 251.21681134298956, "package-0": 138.79266938463735, "core": 0.0, "uncore": 0.0, "dram": 39.16294168329968}, "watts": {"psys": 21.32861957771385, "package-0": 10.760797328028623, "core": 2.9017695432771347, "uncore": 0.0, "dram": 2.2753194256885036}, "steady_cv": 0.019993388442811295, "gpu_mhz": 0.0, "npu_util": 0.8542714285714285, "int8_kernel_seen": false, "int8_verified": null, "n_nodes": 14, "summed_node_us": 2135.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["low_signal:core"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "decode", "seq": 512, "compile_s": 1.6524126464501023, "compile_path": "npuw", "flops": 135276544, "bytes": 135343104, "intensity": 0.9995082128454805, "tflops": 0.045308624177682576, "gbps": 45.33091734053324, "j_igpu": 0.0, "j_npu": 0.013484196597773827, "j_cpu": 0.0, "j_dram": 0.005300434089188764, "j_package": 0.01878463068696259, "j_psys": 0.03400046302414261, "watts_psys": 11.387888509541392, "gflops_per_watt": 3.9786676994352863}
{"device": "NPU", "precision": "int8", "flop_per_rep": 81077993472, "bytes_per_rep": 138416128, "one_s": 0.03009451282698484, "reps_list": [14, 28, 42], "windows": [{"reps": 14, "dt": 0.43416191171854734, "joules": {"package-0": 4.943652, "core": 1.561703, "uncore": 0.0, "dram": 0.510497, "psys": 9.912817}}, {"reps": 28, "dt": 0.842210833914578, "joules": {"package-0": 9.848729, "core": 2.961235, "uncore": 0.0, "dram": 1.008542, "psys": 19.301648}}, {"reps": 42, "dt": 1.2639695387333632, "joules": {"package-0": 15.075828, "core": 4.757922, "uncore": 0.0, "dram": 1.515621, "psys": 29.188646}}], "slope_j": {"psys": 0.6884224642857145, "package-0": 0.36186342857142845, "core": 0.11415067857142842, "uncore": 0.0, "dram": 0.03589728571428576}, "intercept_j": {"psys": 0.19187466666665878, "package-0": -0.1761063333333297, "core": -0.10259899999999551, "uncore": 0.0, "dram": 0.006429333333332139}, "r2": {"psys": 0.9997774098995237, "package-0": 0.9996634123488671, "core": 0.9948796892229911, "uncore": 1.0, "dram": 0.9999730730020525}, "net_j_per_op": {"psys": 0.4250483992646257, "package-0": 0.21969217798826454, "core": 0.028990645044150276, "uncore": 0.0, "dram": 0.0214566253571788}, "pJ_per_flop": {"psys": 5.242463226614194, "package-0": 2.709640046335563, "core": 0.3575649051325141, "uncore": 0.0, "dram": 0.26464179043341485}, "pJ_per_byte": {"psys": 3070.801108268436, "package-0": 1587.1862705787041, "core": 209.4455715749416, "uncore": 0.0, "dram": 155.01535599362236}, "watts": {"psys": 23.092839744579795, "package-0": 11.927366552763322, "core": 3.7642695129883923, "uncore": 0.0, "dram": 1.1990961439773458}, "steady_cv": 0.052951823092962404, "gpu_mhz": 0.0, "npu_util": 0.9483840000000001, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 40, "summed_node_us": 10017.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["overhead_bound:0.33"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "prefill", "seq": 512, "compile_s": 4.130600237287581, "compile_path": "npuw", "flops": 81077993472, "bytes": 138416128, "intensity": 585.7553931287545, "tflops": 2.6941121771308363, "gbps": 4.599380916905439, "j_igpu": 0.0, "j_npu": 0.16924490758693547, "j_cpu": 0.028990645044150276, "j_dram": 0.0214566253571788, "j_package": 0.21969217798826454, "j_psys": 0.4250483992646257, "watts_psys": 14.123784016981915, "gflops_per_watt": 190.7500266140813}
{"device": "NPU", "precision": "int8", "flop_per_rep": 11816402944, "bytes_per_rep": 33556480, "one_s": 0.01830832436680794, "reps_list": [20, 40, 60], "windows": [{"reps": 20, "dt": 0.3855401622131467, "joules": {"package-0": 3.343131, "core": 1.320126, "uncore": 0.0, "dram": 0.389159, "psys": 7.166852}}, {"reps": 40, "dt": 0.7323329746723175, "joules": {"package-0": 6.921735, "core": 2.675775, "uncore": 0.0, "dram": 0.749693, "psys": 14.333154}}, {"reps": 60, "dt": 1.0984652629122138, "joules": {"package-0": 10.512851, "core": 4.153188, "uncore": 0.0, "dram": 1.123227, "psys": 21.524115}}], "slope_j": {"psys": 0.3589315749999999, "package-0": 0.17924299999999999, "core": 0.07082654999999997, "uncore": 0.0, "dram": 0.018351700000000002}, "intercept_j": {"psys": -0.01588933333332913, "package-0": -0.2438143333333329, "core": -0.11669899999999878, "uncore": 0.0, "dram": 0.0199583333333333}, "r2": {"psys": 0.9999990167002528, "package-0": 0.9999989848566244, "core": 0.9993846289433871, "uncore": 1.0, "dram": 0.999895468431945}, "net_j_per_op": {"psys": 0.19870509650764254, "package-0": 0.0927515726000741, "core": 0.019018516892130022, "uncore": 0.0, "dram": 0.00956656712847688}, "pJ_per_flop": {"psys": 16.816039318339154, "package-0": 7.84939148060878, "core": 1.6095013839881815, "uncore": 0.0, "dram": 0.8096006182096628}, "pJ_per_byte": {"psys": 5921.511925793246, "package-0": 2764.0435647622785, "core": 566.7613793857407, "uncore": 0.0, "dram": 285.0885172841991}, "watts": {"psys": 19.59471612505615, "package-0": 9.57049016928281, "core": 3.780900625832454, "uncore": 0.0, "dram": 1.0225421211974774}, "steady_cv": 0.019933447867327717, "gpu_mhz": 0.0, "npu_util": 0.9298666666666667, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 24, "summed_node_us": 2599.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["overhead_bound:0.14"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "prefill", "seq": 512, "compile_s": 1.301027724519372, "compile_path": "npuw", "flops": 11816402944, "bytes": 33556480, "intensity": 352.134757400061, "tflops": 0.645411491912528, "gbps": 1.8328536969137488, "j_igpu": 0.0, "j_npu": 0.0641664885794672, "j_cpu": 0.019018516892130022, "j_dram": 0.00956656712847688, "j_package": 0.0927515726000741, "j_psys": 0.19870509650764254, "watts_psys": 10.853265024509003, "gflops_per_watt": 59.4670350770068}
{"device": "NPU", "precision": "int8", "flop_per_rep": 69261590528, "bytes_per_rep": 104859648, "one_s": 0.023517012363299727, "reps_list": [16, 32, 48], "windows": [{"reps": 16, "dt": 0.3910400988534093, "joules": {"package-0": 4.696216, "core": 1.336606, "uncore": 0.0, "dram": 0.515013, "psys": 9.359717}}, {"reps": 32, "dt": 0.7525443956255913, "joules": {"package-0": 9.678564, "core": 2.596124, "uncore": 0.0, "dram": 1.005796, "psys": 18.684095}}, {"reps": 48, "dt": 1.1257009077817202, "joules": {"package-0": 14.485253, "core": 3.880971, "uncore": 0.0, "dram": 1.506405, "psys": 27.676626}}], "slope_j": {"psys": 0.5724034062499991, "package-0": 0.30590740624999996, "core": 0.07951140624999997, "uncore": 0.0, "dram": 0.030981000000000036}, "intercept_j": {"psys": 0.25657033333336204, "package-0": -0.1690259999999976, "core": 0.0602020000000009, "uncore": 0.0, "dram": 0.01767933333333212}, "r2": {"psys": 0.9998906036962251, "package-0": 0.9998926769503269, "core": 0.9999669674834304, "uncore": 1.0, "dram": 0.9999672563398453}, "net_j_per_op": {"psys": 0.3665927596752625, "package-0": 0.1948093108300241, "core": 0.012964073424454103, "uncore": 0.0, "dram": 0.0196965113590081}, "pJ_per_flop": {"psys": 5.292872382522924, "package-0": 2.812660081077255, "core": 0.18717550846905817, "uncore": 0.0, "dram": 0.28437855973067067}, "pJ_per_byte": {"psys": 3496.032712938942, "package-0": 1857.8100780008729, "core": 123.63262390938125, "uncore": 0.0, "dram": 187.83690136942002}, "watts": {"psys": 24.586127459503352, "package-0": 12.867763452855609, "core": 3.4476040422209047, "uncore": 0.0, "dram": 1.3381929334750973}, "steady_cv": 0.020688361110441908, "gpu_mhz": 0.0, "npu_util": 0.9531272727272728, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 18, "summed_node_us": 8055.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["overhead_bound:0.34"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "prefill", "seq": 512, "compile_s": 3.553360056132078, "compile_path": "npuw", "flops": 69261590528, "bytes": 104859648, "intensity": 660.5170992754048, "tflops": 2.945169626907563, "gbps": 4.458884758832813, "j_igpu": 0.0, "j_npu": 0.1621487260465619, "j_cpu": 0.012964073424454103, "j_dram": 0.0196965113590081, "j_package": 0.1948093108300241, "j_psys": 0.3665927596752625, "watts_psys": 15.58840698010438, "gflops_per_watt": 188.93332915072773}
{"device": "NPU", "precision": "int8", "flop_per_rep": 158363648, "bytes_per_rep": 77718528, "one_s": 0.0026429597580549763, "reps_list": [136, 272, 408], "windows": [{"reps": 136, "dt": 0.37999118398875, "joules": {"package-0": 3.603018, "core": 1.423275, "uncore": 0.0, "dram": 0.631834, "psys": 7.764995}}, {"reps": 272, "dt": 0.7141607068479061, "joules": {"package-0": 7.464092, "core": 2.627495, "uncore": 0.0, "dram": 1.233334, "psys": 15.487265}}, {"reps": 408, "dt": 1.0783275812864304, "joules": {"package-0": 11.195284, "core": 3.946707, "uncore": 0.0, "dram": 1.850093, "psys": 23.011599}}], "slope_j": {"psys": 0.05605369117647054, "package-0": 0.027912742647058796, "core": 0.009277323529411763, "uncore": 0.0, "dram": 0.0044788933823529465}, "intercept_j": {"psys": 0.17468233333334618, "package-0": -0.17146799999999254, "core": 0.14239366666666697, "uncore": 0.0, "dram": 0.02016133333333192}, "r2": {"psys": 0.9999438231109392, "package-0": 0.9999024580383462, "core": 0.9993082800936567, "uncore": 1.0, "dram": 0.9999477087619869}, "net_j_per_op": {"psys": 0.032923658829114105, "package-0": 0.015426981664520941, "core": 0.001798400614089101, "uncore": 0.0, "dram": 0.0032106859711550115}, "pJ_per_flop": {"psys": 207.89909328884684, "package-0": 97.41491724490295, "core": 11.35614540837744, "uncore": 0.0, "dram": 20.274134952707147}, "pJ_per_byte": {"psys": 423.626896653448, "package-0": 198.49811957994032, "core": 23.139921205006623, "uncore": 0.0, "dram": 41.31171875971469}, "watts": {"psys": 21.34008199303172, "package-0": 10.382080727865809, "core": 3.6600260148141914, "uncore": 0.0, "dram": 1.7157059061707982}, "steady_cv": 0.015371945077788395, "gpu_mhz": 0.0, "npu_util": 0.739352380952381, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 0, "summed_node_us": null, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "full", "stage": "decode", "seq": 512, "compile_s": 2.7381785167381167, "compile_path": "npuw", "flops": 158363648, "bytes": 77718528, "intensity": 2.0376562973503565, "tflops": 0.059919053824922394, "gbps": 29.40586884198158, "j_igpu": 0.0, "j_npu": 0.010417895079276828, "j_cpu": 0.001798400614089101, "j_dram": 0.0032106859711550115, "j_package": 0.015426981664520941, "j_psys": 0.032923658829114105, "watts_psys": 12.457116960926978, "gflops_per_watt": 4.810025787898166}
{"device": "NPU", "precision": "int8", "flop_per_rep": 23087104, "bytes_per_rep": 10010624, "one_s": 0.001322611988511648, "reps_list": [267, 534, 801], "windows": [{"reps": 267, "dt": 0.37253148667514324, "joules": {"package-0": 2.994743, "core": 1.471005, "uncore": 0.0, "dram": 0.357848, "psys": 6.537276}}, {"reps": 534, "dt": 0.7062748018652201, "joules": {"package-0": 6.290634, "core": 2.821526, "uncore": 0.0, "dram": 0.692931, "psys": 13.336331}}, {"reps": 801, "dt": 1.0267996340990067, "joules": {"package-0": 9.148597, "core": 4.021291, "uncore": 0.0, "dram": 1.02319, "psys": 19.285839}}], "slope_j": {"psys": 0.023873713483146024, "package-0": 0.011524071161048692, "core": 0.004775816479400746, "uncore": 0.0, "dram": 0.0012459588014981278}, "intercept_j": {"psys": 0.3045856666666917, "package-0": -0.009196000000001684, "core": 0.22098800000000166, "uncore": 0.0, "dram": 0.025980999999999737}, "r2": {"psys": 0.9985219514002249, "package-0": 0.9983147773185012, "core": 0.9988365583344113, "uncore": 1.0, "dram": 0.9999824775086251}, "net_j_per_op": {"psys": 0.012298789580487236, "package-0": 0.0052758424050616945, "core": 0.0010331514316968757, "uncore": 0.0, "dram": 0.000611311860371747}, "pJ_per_flop": {"psys": 532.7125299252447, "package-0": 228.5190210544248, "core": 44.750152799453566, "uncore": 0.0, "dram": 26.47849900843982}, "pJ_per_byte": {"psys": 1228.5737213271855, "package-0": 527.0243298581281, "core": 103.20549764898529, "uncore": 0.0, "dram": 61.06630919029093}, "watts": {"psys": 18.782475528366238, "package-0": 8.90981716021713, "core": 3.9163346639956598, "uncore": 0.0, "dram": 0.9964845779262729}, "steady_cv": 0.011359899934840927, "gpu_mhz": 0.0, "npu_util": 0.68491, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 22, "summed_node_us": 642.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": ["overhead_bound:0.49"], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_ffn", "stage": "decode", "seq": 512, "compile_s": 1.157390533015132, "compile_path": "npuw", "flops": 23087104, "bytes": 10010624, "intensity": 2.3062602291325693, "tflops": 0.01745568934845374, "gbps": 7.5688290193597005, "j_igpu": 0.0, "j_npu": 0.003631379112993072, "j_cpu": 0.0010331514316968757, "j_dram": 0.000611311860371747, "j_package": 0.0052758424050616945, "j_psys": 0.012298789580487236, "watts_psys": 9.29886443440394, "gflops_per_watt": 1.8771850554000098}
{"device": "NPU", "precision": "int8", "flop_per_rep": 135276544, "bytes_per_rep": 67707904, "one_s": 0.0019637623468867545, "reps_list": [193, 386, 579], "windows": [{"reps": 193, "dt": 0.39970810525119305, "joules": {"package-0": 3.967397, "core": 1.622616, "uncore": 0.0, "dram": 0.726622, "psys": 8.152018}}, {"reps": 386, "dt": 0.7580122658982873, "joules": {"package-0": 8.20109, "core": 3.018669, "uncore": 0.0, "dram": 1.425839, "psys": 16.415241}}, {"reps": 579, "dt": 1.0961280465126038, "joules": {"package-0": 12.087128, "core": 4.446583, "uncore": 0.0, "dram": 2.116145, "psys": 23.832702}}], "slope_j": {"psys": 0.040623533678756506, "package-0": 0.02103557253886009, "core": 0.007315976683937829, "uncore": 0.0, "dram": 0.003599800518134713}, "intercept_j": {"psys": 0.45263633333332365, "package-0": -0.03452599999999606, "core": 0.20532233333333139, "uncore": 0.0, "dram": 0.03334566666666768}, "r2": {"psys": 0.9990312222905955, "package-0": 0.999389301128553, "core": 0.9999575712840355, "uncore": 1.0, "dram": 0.999986291365438}, "net_j_per_op": {"psys": 0.0234375411688144, "package-0": 0.011758447824793374, "core": 0.001759014580516481, "uncore": 0.0, "dram": 0.002657501659984097}, "pJ_per_flop": {"psys": 173.2565046074388, "package-0": 86.92155696107505, "core": 13.003101117932765, "uncore": 0.0, "dram": 19.644955299745806}, "pJ_per_byte": {"psys": 346.1566491382512, "package-0": 173.66433060449447, "core": 25.97945699982798, "uncore": 0.0, "dram": 39.24950416400568}, "watts": {"psys": 21.742625850898673, "package-0": 11.027113153847228, "core": 4.056627338518585, "uncore": 0.0, "dram": 1.9305636843547982}, "steady_cv": 0.012243134769324239, "gpu_mhz": 0.0, "npu_util": 0.7733714285714286, "int8_kernel_seen": false, "int8_verified": true, "n_nodes": 16, "summed_node_us": 1070.0, "unsupported_ops": [], "fallback": false, "fallback_reasons": [], "flags": [], "model_key": "qwen25_3b_int8", "unit": "block", "part": "no_attn", "stage": "decode", "seq": 512, "compile_s": 2.454532985575497, "compile_path": "npuw", "flops": 135276544, "bytes": 67707904, "intensity": 1.99794316480392, "tflops": 0.06888641296868753, "gbps": 34.47866495013542, "j_igpu": 0.0, "j_npu": 0.007341931584292796, "j_cpu": 0.001759014580516481, "j_dram": 0.002657501659984097, "j_package": 0.011758447824793374, "j_psys": 0.0234375411688144, "watts_psys": 11.935019125899345, "gflops_per_watt": 5.771789072310909}
