{ "algorithm": "activation_calibrated_block_gptq_mxfp4_cuda", "scope": "exact 32x32 Hessian and GPTQ error propagation within each MXFP4 group", "source_model": "jinaai/jina-code-embeddings-1.5b", "source_revision": "39aeb4fb9b60f930934c78ae5d749a46287c248a", "corpus_sha256": "3bc24ff6718944f15be98a93785fa04de388fe6fe45e3101ebe6e40d0abd1d6f", "calibration": { "samples": 64, "sequence_length": 512, "tokens": 32768, "tokens_sha256": "1594ce4891fc875a15b5c0126b429ceeab9bd602c3a0d2c736d50ff69c765315", "damp": 0.01, "scale_offsets": [ 0, -1, -2, -3, -4, 1, 2 ] }, "weighted_error_rtn": 272101819.2921753, "weighted_error_calibrated": 156526689.23944092, "elapsed_seconds": 206.717322807, "cuda": { "device": "NVIDIA RTX A6000", "torch": "2.13.0+cu130", "peak_memory_bytes": 3456967680 }, "layers": [ { "path": "layers.0.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 1, "layer_count": 196, "elapsed_seconds": 0.8038748559999931, "weighted_error_rtn": 820099.4384765625, "weighted_error_gptq_base_scale": 678314.255859375, "weighted_error_calibrated": 449373.44140625, "improvement_vs_rtn": 0.4520500559773367 }, { "path": "layers.0.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 2, "layer_count": 196, "elapsed_seconds": 0.08329766099998892, "weighted_error_rtn": 106296.73828125, "weighted_error_gptq_base_scale": 86446.89453125, "weighted_error_calibrated": 70645.796875, "improvement_vs_rtn": 0.33539073712611345 }, { "path": "layers.0.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 3, "layer_count": 196, "elapsed_seconds": 0.08253572899999995, "weighted_error_rtn": 12436.029296875, "weighted_error_gptq_base_scale": 9718.24755859375, "weighted_error_calibrated": 7840.267578125, "improvement_vs_rtn": 0.3695521785160839 }, { "path": "layers.0.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 4, "layer_count": 196, "elapsed_seconds": 0.5074308439999982, "weighted_error_rtn": 16534.144287109375, "weighted_error_gptq_base_scale": 13835.000061035156, "weighted_error_calibrated": 11595.346069335938, "improvement_vs_rtn": 0.2987029828706591 }, { "path": "layers.0.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 5, "layer_count": 196, "elapsed_seconds": 2.981520113000016, "weighted_error_rtn": 581172.8720703125, "weighted_error_gptq_base_scale": 510989.96875, "weighted_error_calibrated": 400968.75439453125, "improvement_vs_rtn": 0.310069733698753 }, { "path": "layers.0.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 6, "layer_count": 196, "elapsed_seconds": 2.9449673459999985, "weighted_error_rtn": 281785.7614746094, "weighted_error_gptq_base_scale": 263538.021484375, "weighted_error_calibrated": 232304.6533203125, "improvement_vs_rtn": 0.17559832652777752 }, { "path": "layers.0.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 7, "layer_count": 196, "elapsed_seconds": 0.5237822520000179, "weighted_error_rtn": 19962.126342773438, "weighted_error_gptq_base_scale": 19675.976928710938, "weighted_error_calibrated": 17210.513427734375, "improvement_vs_rtn": 0.1378416741679016 }, { "path": "layers.1.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 8, "layer_count": 196, "elapsed_seconds": 0.4984646810000015, "weighted_error_rtn": 202926.8203125, "weighted_error_gptq_base_scale": 172663.2021484375, "weighted_error_calibrated": 122042.21435546875, "improvement_vs_rtn": 0.3985900229081196 }, { "path": "layers.1.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 9, "layer_count": 196, "elapsed_seconds": 0.09255734100000268, "weighted_error_rtn": 53464.1337890625, "weighted_error_gptq_base_scale": 44969.7158203125, "weighted_error_calibrated": 33074.40234375, "improvement_vs_rtn": 0.38137214615237547 }, { "path": "layers.1.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 10, "layer_count": 196, "elapsed_seconds": 0.08957092800000055, "weighted_error_rtn": 7117.267333984375, "weighted_error_gptq_base_scale": 6516.740234375, "weighted_error_calibrated": 5391.9306640625, "improvement_vs_rtn": 0.24241560545063867 }, { "path": "layers.1.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 11, "layer_count": 196, "elapsed_seconds": 0.49742364299999053, "weighted_error_rtn": 9950.6337890625, "weighted_error_gptq_base_scale": 7533.7913818359375, "weighted_error_calibrated": 6307.647888183594, "improvement_vs_rtn": 0.36610591627672895 }, { "path": "layers.1.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 12, "layer_count": 196, "elapsed_seconds": 2.9774189229999877, "weighted_error_rtn": 28107432.59375, "weighted_error_gptq_base_scale": 22609605.28125, "weighted_error_calibrated": 18853021.40625, "improvement_vs_rtn": 0.32925138774709084 }, { "path": "layers.1.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 13, "layer_count": 196, "elapsed_seconds": 2.871950090000013, "weighted_error_rtn": 13041386.859375, "weighted_error_gptq_base_scale": 11148947.625, "weighted_error_calibrated": 6689813.7109375, "improvement_vs_rtn": 0.4870320324767894 }, { "path": "layers.1.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 14, "layer_count": 196, "elapsed_seconds": 0.49141874500000426, "weighted_error_rtn": 764725.7568359375, "weighted_error_gptq_base_scale": 759056.822265625, "weighted_error_calibrated": 456867.82373046875, "improvement_vs_rtn": 0.40257298822944687 }, { "path": "layers.2.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 15, "layer_count": 196, "elapsed_seconds": 0.49422986300001526, "weighted_error_rtn": 398811.712890625, "weighted_error_gptq_base_scale": 343660.2734375, "weighted_error_calibrated": 265794.9150390625, "improvement_vs_rtn": 0.33353282652468796 }, { "path": "layers.2.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 16, "layer_count": 196, "elapsed_seconds": 0.08133687099999065, "weighted_error_rtn": 98592.0546875, "weighted_error_gptq_base_scale": 83920.18359375, "weighted_error_calibrated": 63935.69140625, "improvement_vs_rtn": 0.35151273995757293 }, { "path": "layers.2.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 17, "layer_count": 196, "elapsed_seconds": 0.08125429700001519, "weighted_error_rtn": 20962.69140625, "weighted_error_gptq_base_scale": 19174.521484375, "weighted_error_calibrated": 16984.3408203125, "improvement_vs_rtn": 0.18978243341174028 }, { "path": "layers.2.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 18, "layer_count": 196, "elapsed_seconds": 0.48289627799999835, "weighted_error_rtn": 12438.599792480469, "weighted_error_gptq_base_scale": 9747.996276855469, "weighted_error_calibrated": 8307.598266601562, "improvement_vs_rtn": 0.3321114590708376 }, { "path": "layers.2.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 19, "layer_count": 196, "elapsed_seconds": 2.922003638999996, "weighted_error_rtn": 20588880.90625, "weighted_error_gptq_base_scale": 15993459.09375, "weighted_error_calibrated": 12410532.03125, "improvement_vs_rtn": 0.3972216320177637 }, { "path": "layers.2.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 20, "layer_count": 196, "elapsed_seconds": 2.8585549720000074, "weighted_error_rtn": 4559173.76171875, "weighted_error_gptq_base_scale": 3871353.20703125, "weighted_error_calibrated": 3279228.9765625, "improvement_vs_rtn": 0.2807405139728053 }, { "path": "layers.2.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 21, "layer_count": 196, "elapsed_seconds": 0.4994389250000211, "weighted_error_rtn": 9352489.0546875, "weighted_error_gptq_base_scale": 9347508.0625, "weighted_error_calibrated": 3744747.16015625, "improvement_vs_rtn": 0.5995988727429336 }, { "path": "layers.3.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 22, "layer_count": 196, "elapsed_seconds": 0.5087138780000089, "weighted_error_rtn": 427570.7734375, "weighted_error_gptq_base_scale": 359414.63671875, "weighted_error_calibrated": 276979.986328125, "improvement_vs_rtn": 0.35220084361399306 }, { "path": "layers.3.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 23, "layer_count": 196, "elapsed_seconds": 0.08266724699998917, "weighted_error_rtn": 108658.359375, "weighted_error_gptq_base_scale": 91839.51171875, "weighted_error_calibrated": 68450.26171875, "improvement_vs_rtn": 0.37004145734875726 }, { "path": "layers.3.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 24, "layer_count": 196, "elapsed_seconds": 0.08282572499999219, "weighted_error_rtn": 29908.263671875, "weighted_error_gptq_base_scale": 27178.3857421875, "weighted_error_calibrated": 23691.4140625, "improvement_vs_rtn": 0.20786394280792608 }, { "path": "layers.3.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 25, "layer_count": 196, "elapsed_seconds": 0.4879630529999872, "weighted_error_rtn": 15724.072387695312, "weighted_error_gptq_base_scale": 11875.04052734375, "weighted_error_calibrated": 10161.631591796875, "improvement_vs_rtn": 0.35375319184178144 }, { "path": "layers.3.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 26, "layer_count": 196, "elapsed_seconds": 2.876083354000002, "weighted_error_rtn": 21423931.953125, "weighted_error_gptq_base_scale": 15982921.21875, "weighted_error_calibrated": 10569723.8984375, "improvement_vs_rtn": 0.5066394011349654 }, { "path": "layers.3.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 27, "layer_count": 196, "elapsed_seconds": 2.847224725999979, "weighted_error_rtn": 2739761.05078125, "weighted_error_gptq_base_scale": 2340602.67578125, "weighted_error_calibrated": 1992757.88671875, "improvement_vs_rtn": 0.2726526694178275 }, { "path": "layers.3.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 28, "layer_count": 196, "elapsed_seconds": 0.4886976139999888, "weighted_error_rtn": 201920.28955078125, "weighted_error_gptq_base_scale": 199625.76904296875, "weighted_error_calibrated": 64935.53759765625, "improvement_vs_rtn": 0.6784100411993244 }, { "path": "layers.4.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 29, "layer_count": 196, "elapsed_seconds": 0.48968343499998923, "weighted_error_rtn": 506273.9765625, "weighted_error_gptq_base_scale": 423747.0, "weighted_error_calibrated": 327162.8125, "improvement_vs_rtn": 0.3537830746874041 }, { "path": "layers.4.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 30, "layer_count": 196, "elapsed_seconds": 0.08045120400001338, "weighted_error_rtn": 117847.046875, "weighted_error_gptq_base_scale": 99696.51953125, "weighted_error_calibrated": 76436.05859375, "improvement_vs_rtn": 0.3513960627725743 }, { "path": "layers.4.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 31, "layer_count": 196, "elapsed_seconds": 0.0805687700000135, "weighted_error_rtn": 36880.6015625, "weighted_error_gptq_base_scale": 33629.2919921875, "weighted_error_calibrated": 30236.5234375, "improvement_vs_rtn": 0.18015102366865032 }, { "path": "layers.4.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 32, "layer_count": 196, "elapsed_seconds": 0.478373409999989, "weighted_error_rtn": 16774.255859375, "weighted_error_gptq_base_scale": 13126.910400390625, "weighted_error_calibrated": 11206.513488769531, "improvement_vs_rtn": 0.3319218698750026 }, { "path": "layers.4.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 33, "layer_count": 196, "elapsed_seconds": 2.873471132999981, "weighted_error_rtn": 10407971.8125, "weighted_error_gptq_base_scale": 7780122.2421875, "weighted_error_calibrated": 4910707.8125, "improvement_vs_rtn": 0.5281782175272394 }, { "path": "layers.4.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 34, "layer_count": 196, "elapsed_seconds": 2.848669927000003, "weighted_error_rtn": 1656545.568359375, "weighted_error_gptq_base_scale": 1476120.6171875, "weighted_error_calibrated": 1318684.08984375, "improvement_vs_rtn": 0.20395543893805435 }, { "path": "layers.4.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 35, "layer_count": 196, "elapsed_seconds": 0.4923134969999978, "weighted_error_rtn": 148637.029296875, "weighted_error_gptq_base_scale": 146750.43017578125, "weighted_error_calibrated": 66460.77099609375, "improvement_vs_rtn": 0.5528653168696567 }, { "path": "layers.5.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 36, "layer_count": 196, "elapsed_seconds": 0.49192502200000376, "weighted_error_rtn": 506808.310546875, "weighted_error_gptq_base_scale": 429508.564453125, "weighted_error_calibrated": 321047.3671875, "improvement_vs_rtn": 0.36653097333571416 }, { "path": "layers.5.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 37, "layer_count": 196, "elapsed_seconds": 0.08120123800000556, "weighted_error_rtn": 124024.56640625, "weighted_error_gptq_base_scale": 107803.6484375, "weighted_error_calibrated": 79536.078125, "improvement_vs_rtn": 0.3587070656270247 }, { "path": "layers.5.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 38, "layer_count": 196, "elapsed_seconds": 0.08118492399998445, "weighted_error_rtn": 43522.78515625, "weighted_error_gptq_base_scale": 39195.0859375, "weighted_error_calibrated": 34612.45703125, "improvement_vs_rtn": 0.2047278935162643 }, { "path": "layers.5.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 39, "layer_count": 196, "elapsed_seconds": 0.4812568130000159, "weighted_error_rtn": 25490.9541015625, "weighted_error_gptq_base_scale": 19351.019409179688, "weighted_error_calibrated": 15785.832153320312, "improvement_vs_rtn": 0.3807280774809171 }, { "path": "layers.5.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 40, "layer_count": 196, "elapsed_seconds": 2.8559198519999995, "weighted_error_rtn": 21155776.4375, "weighted_error_gptq_base_scale": 16139754.765625, "weighted_error_calibrated": 10799717.59375, "improvement_vs_rtn": 0.48951447725611275 }, { "path": "layers.5.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 41, "layer_count": 196, "elapsed_seconds": 2.891749764000025, "weighted_error_rtn": 3951933.88671875, "weighted_error_gptq_base_scale": 3438590.86328125, "weighted_error_calibrated": 3007535.38671875, "improvement_vs_rtn": 0.23897122954759864 }, { "path": "layers.5.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 42, "layer_count": 196, "elapsed_seconds": 0.4893558499999813, "weighted_error_rtn": 50782.3681640625, "weighted_error_gptq_base_scale": 44689.361328125, "weighted_error_calibrated": 37780.351806640625, "improvement_vs_rtn": 0.2560340690575179 }, { "path": "layers.6.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 43, "layer_count": 196, "elapsed_seconds": 0.4942097719999765, "weighted_error_rtn": 659498.69140625, "weighted_error_gptq_base_scale": 569176.470703125, "weighted_error_calibrated": 397090.958984375, "improvement_vs_rtn": 0.39788969385571127 }, { "path": "layers.6.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 44, "layer_count": 196, "elapsed_seconds": 0.08056902300000957, "weighted_error_rtn": 163575.46875, "weighted_error_gptq_base_scale": 145787.73828125, "weighted_error_calibrated": 101065.21484375, "improvement_vs_rtn": 0.3821493185007303 }, { "path": "layers.6.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 45, "layer_count": 196, "elapsed_seconds": 0.08068569299996398, "weighted_error_rtn": 33393.9345703125, "weighted_error_gptq_base_scale": 29326.8916015625, "weighted_error_calibrated": 25563.140625, "improvement_vs_rtn": 0.2344974932146553 }, { "path": "layers.6.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 46, "layer_count": 196, "elapsed_seconds": 0.48161964499996657, "weighted_error_rtn": 16923.45166015625, "weighted_error_gptq_base_scale": 11837.577514648438, "weighted_error_calibrated": 9883.885925292969, "improvement_vs_rtn": 0.41596512793172635 }, { "path": "layers.6.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 47, "layer_count": 196, "elapsed_seconds": 2.845994502999986, "weighted_error_rtn": 1975343.072265625, "weighted_error_gptq_base_scale": 1717484.3046875, "weighted_error_calibrated": 1383940.705078125, "improvement_vs_rtn": 0.29939222988196657 }, { "path": "layers.6.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 48, "layer_count": 196, "elapsed_seconds": 2.8382487050000123, "weighted_error_rtn": 1145541.654296875, "weighted_error_gptq_base_scale": 1058650.28515625, "weighted_error_calibrated": 943826.505859375, "improvement_vs_rtn": 0.17608713544450838 }, { "path": "layers.6.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 49, "layer_count": 196, "elapsed_seconds": 0.49043685199995934, "weighted_error_rtn": 64442.0048828125, "weighted_error_gptq_base_scale": 62899.8740234375, "weighted_error_calibrated": 54787.779296875, "improvement_vs_rtn": 0.14981261994398942 }, { "path": "layers.7.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 50, "layer_count": 196, "elapsed_seconds": 0.503478097000027, "weighted_error_rtn": 332239.7177734375, "weighted_error_gptq_base_scale": 283884.66552734375, "weighted_error_calibrated": 218239.056640625, "improvement_vs_rtn": 0.3431277328815707 }, { "path": "layers.7.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 51, "layer_count": 196, "elapsed_seconds": 0.0907050159999585, "weighted_error_rtn": 77428.74609375, "weighted_error_gptq_base_scale": 66312.501953125, "weighted_error_calibrated": 51934.884765625, "improvement_vs_rtn": 0.3292557688750025 }, { "path": "layers.7.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 52, "layer_count": 196, "elapsed_seconds": 0.08583037799996873, "weighted_error_rtn": 29457.5517578125, "weighted_error_gptq_base_scale": 26455.7783203125, "weighted_error_calibrated": 22711.3369140625, "improvement_vs_rtn": 0.2290147836865235 }, { "path": "layers.7.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 53, "layer_count": 196, "elapsed_seconds": 0.5030396940000514, "weighted_error_rtn": 30850.751220703125, "weighted_error_gptq_base_scale": 24093.884765625, "weighted_error_calibrated": 20567.847412109375, "improvement_vs_rtn": 0.33331129394648795 }, { "path": "layers.7.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 54, "layer_count": 196, "elapsed_seconds": 2.8569434920000276, "weighted_error_rtn": 1730606.1328125, "weighted_error_gptq_base_scale": 1548644.73828125, "weighted_error_calibrated": 1267828.77734375, "improvement_vs_rtn": 0.26740767104336205 }, { "path": "layers.7.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 55, "layer_count": 196, "elapsed_seconds": 2.921404963000043, "weighted_error_rtn": 1181460.9638671875, "weighted_error_gptq_base_scale": 1104030.8623046875, "weighted_error_calibrated": 974715.1181640625, "improvement_vs_rtn": 0.17499168574000057 }, { "path": "layers.7.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 56, "layer_count": 196, "elapsed_seconds": 0.4924875780000093, "weighted_error_rtn": 71632.375, "weighted_error_gptq_base_scale": 70237.80126953125, "weighted_error_calibrated": 60851.64501953125, "improvement_vs_rtn": 0.15050080330951965 }, { "path": "layers.8.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 57, "layer_count": 196, "elapsed_seconds": 0.4974517800000058, "weighted_error_rtn": 567222.3515625, "weighted_error_gptq_base_scale": 519214.712890625, "weighted_error_calibrated": 405664.978515625, "improvement_vs_rtn": 0.28482194434306174 }, { "path": "layers.8.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 58, "layer_count": 196, "elapsed_seconds": 0.08160947200002511, "weighted_error_rtn": 147669.6640625, "weighted_error_gptq_base_scale": 129802.01953125, "weighted_error_calibrated": 95872.7578125, "improvement_vs_rtn": 0.3507619969127672 }, { "path": "layers.8.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 59, "layer_count": 196, "elapsed_seconds": 0.08082503500003213, "weighted_error_rtn": 40765.404296875, "weighted_error_gptq_base_scale": 37324.669921875, "weighted_error_calibrated": 33141.294921875, "improvement_vs_rtn": 0.1870240098559368 }, { "path": "layers.8.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 60, "layer_count": 196, "elapsed_seconds": 0.4859186730000147, "weighted_error_rtn": 33881.770263671875, "weighted_error_gptq_base_scale": 28342.121826171875, "weighted_error_calibrated": 24135.089599609375, "improvement_vs_rtn": 0.28766739719361467 }, { "path": "layers.8.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 61, "layer_count": 196, "elapsed_seconds": 2.9269025969999802, "weighted_error_rtn": 1805646.19921875, "weighted_error_gptq_base_scale": 1634345.322265625, "weighted_error_calibrated": 1335643.9765625, "improvement_vs_rtn": 0.2602958557770654 }, { "path": "layers.8.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 62, "layer_count": 196, "elapsed_seconds": 2.8634373200000027, "weighted_error_rtn": 1213027.380859375, "weighted_error_gptq_base_scale": 1147199.701171875, "weighted_error_calibrated": 1017486.5986328125, "improvement_vs_rtn": 0.16120063348284086 }, { "path": "layers.8.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 63, "layer_count": 196, "elapsed_seconds": 0.49374770499997567, "weighted_error_rtn": 70629.65771484375, "weighted_error_gptq_base_scale": 69422.39013671875, "weighted_error_calibrated": 59751.623046875, "improvement_vs_rtn": 0.15401511234681498 }, { "path": "layers.9.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 64, "layer_count": 196, "elapsed_seconds": 0.49106665400000793, "weighted_error_rtn": 455126.94921875, "weighted_error_gptq_base_scale": 413513.25, "weighted_error_calibrated": 334874.951171875, "improvement_vs_rtn": 0.2642163867758085 }, { "path": "layers.9.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 65, "layer_count": 196, "elapsed_seconds": 0.08101344300001756, "weighted_error_rtn": 121171.45703125, "weighted_error_gptq_base_scale": 107031.59375, "weighted_error_calibrated": 82370.9296875, "improvement_vs_rtn": 0.3202117750696303 }, { "path": "layers.9.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 66, "layer_count": 196, "elapsed_seconds": 0.08216231200003676, "weighted_error_rtn": 37358.7109375, "weighted_error_gptq_base_scale": 34474.9794921875, "weighted_error_calibrated": 30235.560546875, "improvement_vs_rtn": 0.19066906249901971 }, { "path": "layers.9.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 67, "layer_count": 196, "elapsed_seconds": 0.4818568569999684, "weighted_error_rtn": 34067.748046875, "weighted_error_gptq_base_scale": 29479.548583984375, "weighted_error_calibrated": 25486.38330078125, "improvement_vs_rtn": 0.25189116504813736 }, { "path": "layers.9.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 68, "layer_count": 196, "elapsed_seconds": 2.849806569000009, "weighted_error_rtn": 1620622.888671875, "weighted_error_gptq_base_scale": 1481386.5234375, "weighted_error_calibrated": 1220219.607421875, "improvement_vs_rtn": 0.24706752203045612 }, { "path": "layers.9.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 69, "layer_count": 196, "elapsed_seconds": 2.858186546000013, "weighted_error_rtn": 1183336.677734375, "weighted_error_gptq_base_scale": 1127817.4697265625, "weighted_error_calibrated": 995469.5068359375, "improvement_vs_rtn": 0.15876054079396018 }, { "path": "layers.9.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 70, "layer_count": 196, "elapsed_seconds": 0.491620609999984, "weighted_error_rtn": 67779.9052734375, "weighted_error_gptq_base_scale": 66839.71337890625, "weighted_error_calibrated": 58349.916015625, "improvement_vs_rtn": 0.1391266219651689 }, { "path": "layers.10.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 71, "layer_count": 196, "elapsed_seconds": 0.48849004800001694, "weighted_error_rtn": 458550.96484375, "weighted_error_gptq_base_scale": 425443.36328125, "weighted_error_calibrated": 342081.47265625, "improvement_vs_rtn": 0.2539946507956573 }, { "path": "layers.10.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 72, "layer_count": 196, "elapsed_seconds": 0.08046142099999543, "weighted_error_rtn": 132803.5234375, "weighted_error_gptq_base_scale": 118750.20703125, "weighted_error_calibrated": 85797.3671875, "improvement_vs_rtn": 0.35395262891591905 }, { "path": "layers.10.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 73, "layer_count": 196, "elapsed_seconds": 0.08025076000001263, "weighted_error_rtn": 52459.578125, "weighted_error_gptq_base_scale": 49804.033203125, "weighted_error_calibrated": 44025.150390625, "improvement_vs_rtn": 0.16077955705777036 }, { "path": "layers.10.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 74, "layer_count": 196, "elapsed_seconds": 0.48554535099998475, "weighted_error_rtn": 60381.6220703125, "weighted_error_gptq_base_scale": 53131.2568359375, "weighted_error_calibrated": 45370.46728515625, "improvement_vs_rtn": 0.24860469577442346 }, { "path": "layers.10.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 75, "layer_count": 196, "elapsed_seconds": 2.8784826680000037, "weighted_error_rtn": 1712889.68359375, "weighted_error_gptq_base_scale": 1544273.2578125, "weighted_error_calibrated": 1250355.392578125, "improvement_vs_rtn": 0.27003157030241376 }, { "path": "layers.10.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 76, "layer_count": 196, "elapsed_seconds": 2.8538305800000217, "weighted_error_rtn": 1173726.0439453125, "weighted_error_gptq_base_scale": 1111982.6708984375, "weighted_error_calibrated": 978013.2744140625, "improvement_vs_rtn": 0.16674484692645097 }, { "path": "layers.10.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 77, "layer_count": 196, "elapsed_seconds": 0.49573140300003615, "weighted_error_rtn": 66936.353515625, "weighted_error_gptq_base_scale": 65724.091796875, "weighted_error_calibrated": 57217.765625, "improvement_vs_rtn": 0.14519147488899864 }, { "path": "layers.11.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 78, "layer_count": 196, "elapsed_seconds": 0.49287397400001964, "weighted_error_rtn": 438932.931640625, "weighted_error_gptq_base_scale": 399584.494140625, "weighted_error_calibrated": 320717.0625, "improvement_vs_rtn": 0.26932558625474445 }, { "path": "layers.11.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 79, "layer_count": 196, "elapsed_seconds": 0.08407004299999699, "weighted_error_rtn": 116227.25, "weighted_error_gptq_base_scale": 103185.0546875, "weighted_error_calibrated": 78245.42578125, "improvement_vs_rtn": 0.3267893219425737 }, { "path": "layers.11.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 80, "layer_count": 196, "elapsed_seconds": 0.08473662900001955, "weighted_error_rtn": 44882.828125, "weighted_error_gptq_base_scale": 41419.974609375, "weighted_error_calibrated": 35853.259765625, "improvement_vs_rtn": 0.20118091342701017 }, { "path": "layers.11.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 81, "layer_count": 196, "elapsed_seconds": 0.4876504720000412, "weighted_error_rtn": 56858.5126953125, "weighted_error_gptq_base_scale": 46111.206298828125, "weighted_error_calibrated": 39564.569091796875, "improvement_vs_rtn": 0.30415750929308716 }, { "path": "layers.11.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 82, "layer_count": 196, "elapsed_seconds": 2.851467388000003, "weighted_error_rtn": 1943698.46875, "weighted_error_gptq_base_scale": 1669312.0, "weighted_error_calibrated": 1312977.263671875, "improvement_vs_rtn": 0.32449539638920655 }, { "path": "layers.11.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 83, "layer_count": 196, "elapsed_seconds": 2.8810023329999694, "weighted_error_rtn": 1131605.3388671875, "weighted_error_gptq_base_scale": 1051645.919921875, "weighted_error_calibrated": 918263.775390625, "improvement_vs_rtn": 0.18853000790022045 }, { "path": "layers.11.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 84, "layer_count": 196, "elapsed_seconds": 0.5023647240000173, "weighted_error_rtn": 61847.04443359375, "weighted_error_gptq_base_scale": 60128.0078125, "weighted_error_calibrated": 52328.66357421875, "improvement_vs_rtn": 0.15390195192908618 }, { "path": "layers.12.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 85, "layer_count": 196, "elapsed_seconds": 0.4953242249999903, "weighted_error_rtn": 527563.462890625, "weighted_error_gptq_base_scale": 484331.44921875, "weighted_error_calibrated": 394291.490234375, "improvement_vs_rtn": 0.2526178972403176 }, { "path": "layers.12.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 86, "layer_count": 196, "elapsed_seconds": 0.08068611999999575, "weighted_error_rtn": 138145.765625, "weighted_error_gptq_base_scale": 123614.609375, "weighted_error_calibrated": 94229.9921875, "improvement_vs_rtn": 0.31789445907962477 }, { "path": "layers.12.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 87, "layer_count": 196, "elapsed_seconds": 0.08061687699995446, "weighted_error_rtn": 45186.837890625, "weighted_error_gptq_base_scale": 41780.138671875, "weighted_error_calibrated": 37344.041015625, "improvement_vs_rtn": 0.17356374646049666 }, { "path": "layers.12.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 88, "layer_count": 196, "elapsed_seconds": 0.47991040400000884, "weighted_error_rtn": 33153.74951171875, "weighted_error_gptq_base_scale": 27470.998291015625, "weighted_error_calibrated": 23410.779296875, "improvement_vs_rtn": 0.29387234802506823 }, { "path": "layers.12.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 89, "layer_count": 196, "elapsed_seconds": 2.8782581660000233, "weighted_error_rtn": 1576713.07421875, "weighted_error_gptq_base_scale": 1400650.203125, "weighted_error_calibrated": 1131131.416015625, "improvement_vs_rtn": 0.2826016131209589 }, { "path": "layers.12.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 90, "layer_count": 196, "elapsed_seconds": 2.835638324999991, "weighted_error_rtn": 1087393.951171875, "weighted_error_gptq_base_scale": 1021431.5078125, "weighted_error_calibrated": 897028.548828125, "improvement_vs_rtn": 0.17506571757052247 }, { "path": "layers.12.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 91, "layer_count": 196, "elapsed_seconds": 0.4883852279999701, "weighted_error_rtn": 61401.41552734375, "weighted_error_gptq_base_scale": 60111.15869140625, "weighted_error_calibrated": 51897.96826171875, "improvement_vs_rtn": 0.15477570319845235 }, { "path": "layers.13.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 92, "layer_count": 196, "elapsed_seconds": 0.49587628199998335, "weighted_error_rtn": 378856.373046875, "weighted_error_gptq_base_scale": 341446.294921875, "weighted_error_calibrated": 272200.412109375, "improvement_vs_rtn": 0.28152083091473745 }, { "path": "layers.13.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 93, "layer_count": 196, "elapsed_seconds": 0.08233134099998551, "weighted_error_rtn": 107346.765625, "weighted_error_gptq_base_scale": 92188.70703125, "weighted_error_calibrated": 73925.13671875, "improvement_vs_rtn": 0.31134267261487414 }, { "path": "layers.13.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 94, "layer_count": 196, "elapsed_seconds": 0.08307616600001211, "weighted_error_rtn": 36329.677734375, "weighted_error_gptq_base_scale": 33182.125, "weighted_error_calibrated": 28679.203125, "improvement_vs_rtn": 0.21058470887937852 }, { "path": "layers.13.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 95, "layer_count": 196, "elapsed_seconds": 0.48470126299997673, "weighted_error_rtn": 51463.104736328125, "weighted_error_gptq_base_scale": 44784.984375, "weighted_error_calibrated": 38525.4970703125, "improvement_vs_rtn": 0.25139578601605217 }, { "path": "layers.13.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 96, "layer_count": 196, "elapsed_seconds": 2.8601872159999857, "weighted_error_rtn": 1322830.66015625, "weighted_error_gptq_base_scale": 1174100.091796875, "weighted_error_calibrated": 973403.515625, "improvement_vs_rtn": 0.26415107772749713 }, { "path": "layers.13.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 97, "layer_count": 196, "elapsed_seconds": 2.831794962999993, "weighted_error_rtn": 1031970.0361328125, "weighted_error_gptq_base_scale": 970936.0625, "weighted_error_calibrated": 848856.9208984375, "improvement_vs_rtn": 0.17744034111743212 }, { "path": "layers.13.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 98, "layer_count": 196, "elapsed_seconds": 0.4905409860000418, "weighted_error_rtn": 52272.7783203125, "weighted_error_gptq_base_scale": 51358.50537109375, "weighted_error_calibrated": 44509.423583984375, "improvement_vs_rtn": 0.14851620644222363 }, { "path": "layers.14.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 99, "layer_count": 196, "elapsed_seconds": 0.4943612810000104, "weighted_error_rtn": 739876.2109375, "weighted_error_gptq_base_scale": 687119.517578125, "weighted_error_calibrated": 525045.80078125, "improvement_vs_rtn": 0.2903599372171157 }, { "path": "layers.14.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 100, "layer_count": 196, "elapsed_seconds": 0.08199030899999116, "weighted_error_rtn": 176723.35546875, "weighted_error_gptq_base_scale": 162325.00390625, "weighted_error_calibrated": 113252.4921875, "improvement_vs_rtn": 0.3591537921679716 }, { "path": "layers.14.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 101, "layer_count": 196, "elapsed_seconds": 0.08325984399999697, "weighted_error_rtn": 66651.29296875, "weighted_error_gptq_base_scale": 61818.587890625, "weighted_error_calibrated": 54379.498046875, "improvement_vs_rtn": 0.18411938276469342 }, { "path": "layers.14.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 102, "layer_count": 196, "elapsed_seconds": 0.49411095700003216, "weighted_error_rtn": 43520.869873046875, "weighted_error_gptq_base_scale": 36304.07861328125, "weighted_error_calibrated": 29532.79736328125, "improvement_vs_rtn": 0.32141068297967657 }, { "path": "layers.14.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 103, "layer_count": 196, "elapsed_seconds": 2.840912411999966, "weighted_error_rtn": 1452434.216796875, "weighted_error_gptq_base_scale": 1275502.2978515625, "weighted_error_calibrated": 1040219.0283203125, "improvement_vs_rtn": 0.28380988530113327 }, { "path": "layers.14.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 104, "layer_count": 196, "elapsed_seconds": 2.8455828670000187, "weighted_error_rtn": 1148001.1103515625, "weighted_error_gptq_base_scale": 1080659.4013671875, "weighted_error_calibrated": 947670.462890625, "improvement_vs_rtn": 0.17450387952986257 }, { "path": "layers.14.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 105, "layer_count": 196, "elapsed_seconds": 0.4906151470000282, "weighted_error_rtn": 67093.99462890625, "weighted_error_gptq_base_scale": 66033.17236328125, "weighted_error_calibrated": 55640.318359375, "improvement_vs_rtn": 0.1707109009216249 }, { "path": "layers.15.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 106, "layer_count": 196, "elapsed_seconds": 0.49400441599999567, "weighted_error_rtn": 1371385.689453125, "weighted_error_gptq_base_scale": 1236096.900390625, "weighted_error_calibrated": 895986.578125, "improvement_vs_rtn": 0.3466560246211279 }, { "path": "layers.15.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 107, "layer_count": 196, "elapsed_seconds": 0.08172228099999757, "weighted_error_rtn": 131473.69921875, "weighted_error_gptq_base_scale": 115538.515625, "weighted_error_calibrated": 83631.890625, "improvement_vs_rtn": 0.3638888148583187 }, { "path": "layers.15.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 108, "layer_count": 196, "elapsed_seconds": 0.08135608300000285, "weighted_error_rtn": 57585.345703125, "weighted_error_gptq_base_scale": 53711.76171875, "weighted_error_calibrated": 45800.697265625, "improvement_vs_rtn": 0.2046466560825123 }, { "path": "layers.15.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 109, "layer_count": 196, "elapsed_seconds": 0.4800983050000127, "weighted_error_rtn": 66573.90771484375, "weighted_error_gptq_base_scale": 54760.44140625, "weighted_error_calibrated": 47316.497802734375, "improvement_vs_rtn": 0.2892636255422275 }, { "path": "layers.15.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 110, "layer_count": 196, "elapsed_seconds": 2.8476686390000054, "weighted_error_rtn": 1415045.43359375, "weighted_error_gptq_base_scale": 1172149.0771484375, "weighted_error_calibrated": 940205.88671875, "improvement_vs_rtn": 0.3355648770012025 }, { "path": "layers.15.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 111, "layer_count": 196, "elapsed_seconds": 2.831139108000002, "weighted_error_rtn": 993214.6328125, "weighted_error_gptq_base_scale": 903824.630859375, "weighted_error_calibrated": 789267.388671875, "improvement_vs_rtn": 0.2053405552061841 }, { "path": "layers.15.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 112, "layer_count": 196, "elapsed_seconds": 0.49057252799997286, "weighted_error_rtn": 50093.28173828125, "weighted_error_gptq_base_scale": 48330.845703125, "weighted_error_calibrated": 41418.195068359375, "improvement_vs_rtn": 0.17317864529710736 }, { "path": "layers.16.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 113, "layer_count": 196, "elapsed_seconds": 0.4918157519999795, "weighted_error_rtn": 520895.37109375, "weighted_error_gptq_base_scale": 465019.005859375, "weighted_error_calibrated": 388667.3984375, "improvement_vs_rtn": 0.25384747109309946 }, { "path": "layers.16.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 114, "layer_count": 196, "elapsed_seconds": 0.08070306500002289, "weighted_error_rtn": 158985.1640625, "weighted_error_gptq_base_scale": 135686.3515625, "weighted_error_calibrated": 103121.34765625, "improvement_vs_rtn": 0.3513775435316021 }, { "path": "layers.16.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 115, "layer_count": 196, "elapsed_seconds": 0.08035287699999571, "weighted_error_rtn": 63549.701171875, "weighted_error_gptq_base_scale": 57044.95703125, "weighted_error_calibrated": 49201.76171875, "improvement_vs_rtn": 0.22577508923794787 }, { "path": "layers.16.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 116, "layer_count": 196, "elapsed_seconds": 0.4784325709999848, "weighted_error_rtn": 47129.232177734375, "weighted_error_gptq_base_scale": 34381.3310546875, "weighted_error_calibrated": 29185.79443359375, "improvement_vs_rtn": 0.3807284123889839 }, { "path": "layers.16.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 117, "layer_count": 196, "elapsed_seconds": 2.8727582970000185, "weighted_error_rtn": 1470111.923828125, "weighted_error_gptq_base_scale": 1194082.767578125, "weighted_error_calibrated": 963546.59765625, "improvement_vs_rtn": 0.3445760271454672 }, { "path": "layers.16.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 118, "layer_count": 196, "elapsed_seconds": 2.8393550700000105, "weighted_error_rtn": 1045667.751953125, "weighted_error_gptq_base_scale": 933034.802734375, "weighted_error_calibrated": 816386.380859375, "improvement_vs_rtn": 0.21926789906783717 }, { "path": "layers.16.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 119, "layer_count": 196, "elapsed_seconds": 0.49250315699998737, "weighted_error_rtn": 50047.448974609375, "weighted_error_gptq_base_scale": 47815.63134765625, "weighted_error_calibrated": 40935.44921875, "improvement_vs_rtn": 0.18206721706199602 }, { "path": "layers.17.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 120, "layer_count": 196, "elapsed_seconds": 0.49305278700001054, "weighted_error_rtn": 486840.177734375, "weighted_error_gptq_base_scale": 424480.0244140625, "weighted_error_calibrated": 334938.7890625, "improvement_vs_rtn": 0.31201489856236553 }, { "path": "layers.17.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 121, "layer_count": 196, "elapsed_seconds": 0.08121057199997495, "weighted_error_rtn": 111227.765625, "weighted_error_gptq_base_scale": 91918.796875, "weighted_error_calibrated": 61687.35546875, "improvement_vs_rtn": 0.4453960742434899 }, { "path": "layers.17.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 122, "layer_count": 196, "elapsed_seconds": 0.08096389800005, "weighted_error_rtn": 71060.505859375, "weighted_error_gptq_base_scale": 63553.98046875, "weighted_error_calibrated": 55619.328125, "improvement_vs_rtn": 0.21729619776324527 }, { "path": "layers.17.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 123, "layer_count": 196, "elapsed_seconds": 0.49947484399996256, "weighted_error_rtn": 37273.164794921875, "weighted_error_gptq_base_scale": 23114.548706054688, "weighted_error_calibrated": 19342.155029296875, "improvement_vs_rtn": 0.4810702247657793 }, { "path": "layers.17.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 124, "layer_count": 196, "elapsed_seconds": 2.864739109000027, "weighted_error_rtn": 1616565.8125, "weighted_error_gptq_base_scale": 1244583.97265625, "weighted_error_calibrated": 988522.7060546875, "improvement_vs_rtn": 0.38850450850129714 }, { "path": "layers.17.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 125, "layer_count": 196, "elapsed_seconds": 2.8452946340000267, "weighted_error_rtn": 1084306.6484375, "weighted_error_gptq_base_scale": 925986.6162109375, "weighted_error_calibrated": 808290.6669921875, "improvement_vs_rtn": 0.25455527902835895 }, { "path": "layers.17.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 126, "layer_count": 196, "elapsed_seconds": 0.4907975510000142, "weighted_error_rtn": 60200.4140625, "weighted_error_gptq_base_scale": 56101.26171875, "weighted_error_calibrated": 48240.280517578125, "improvement_vs_rtn": 0.19867194821126777 }, { "path": "layers.18.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 127, "layer_count": 196, "elapsed_seconds": 0.4946676449999927, "weighted_error_rtn": 429544.6171875, "weighted_error_gptq_base_scale": 349587.9208984375, "weighted_error_calibrated": 273729.783203125, "improvement_vs_rtn": 0.3627442359878542 }, { "path": "layers.18.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 128, "layer_count": 196, "elapsed_seconds": 0.08141851999999972, "weighted_error_rtn": 134299.609375, "weighted_error_gptq_base_scale": 104409.51953125, "weighted_error_calibrated": 68598.2265625, "improvement_vs_rtn": 0.489214995622544 }, { "path": "layers.18.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 129, "layer_count": 196, "elapsed_seconds": 0.0808827320000205, "weighted_error_rtn": 52278.740234375, "weighted_error_gptq_base_scale": 43137.482421875, "weighted_error_calibrated": 37407.615234375, "improvement_vs_rtn": 0.28445836554840587 }, { "path": "layers.18.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 130, "layer_count": 196, "elapsed_seconds": 0.49450191599999016, "weighted_error_rtn": 49491.3486328125, "weighted_error_gptq_base_scale": 33781.177734375, "weighted_error_calibrated": 28722.70947265625, "improvement_vs_rtn": 0.4196418108191692 }, { "path": "layers.18.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 131, "layer_count": 196, "elapsed_seconds": 2.8343878599999925, "weighted_error_rtn": 1585543.890625, "weighted_error_gptq_base_scale": 1153030.6416015625, "weighted_error_calibrated": 917661.751953125, "improvement_vs_rtn": 0.42123219837743175 }, { "path": "layers.18.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 132, "layer_count": 196, "elapsed_seconds": 2.8492617499999824, "weighted_error_rtn": 1102573.7099609375, "weighted_error_gptq_base_scale": 887495.552734375, "weighted_error_calibrated": 771507.2978515625, "improvement_vs_rtn": 0.3002669201328083 }, { "path": "layers.18.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 133, "layer_count": 196, "elapsed_seconds": 0.5218132050000008, "weighted_error_rtn": 65883.90673828125, "weighted_error_gptq_base_scale": 59226.7421875, "weighted_error_calibrated": 51242.657470703125, "improvement_vs_rtn": 0.22222800669273246 }, { "path": "layers.19.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 134, "layer_count": 196, "elapsed_seconds": 0.4867803550000076, "weighted_error_rtn": 493562.900390625, "weighted_error_gptq_base_scale": 378295.91015625, "weighted_error_calibrated": 313651.927734375, "improvement_vs_rtn": 0.3645147812241589 }, { "path": "layers.19.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 135, "layer_count": 196, "elapsed_seconds": 0.08143770800000993, "weighted_error_rtn": 132369.65234375, "weighted_error_gptq_base_scale": 95312.6953125, "weighted_error_calibrated": 67747.89453125, "improvement_vs_rtn": 0.48819164112242375 }, { "path": "layers.19.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 136, "layer_count": 196, "elapsed_seconds": 0.08088364799999681, "weighted_error_rtn": 77560.181640625, "weighted_error_gptq_base_scale": 62823.400390625, "weighted_error_calibrated": 54863.015625, "improvement_vs_rtn": 0.2926394128470752 }, { "path": "layers.19.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 137, "layer_count": 196, "elapsed_seconds": 0.4927817679999862, "weighted_error_rtn": 50846.86328125, "weighted_error_gptq_base_scale": 28051.048095703125, "weighted_error_calibrated": 23426.233032226562, "improvement_vs_rtn": 0.5392786984194345 }, { "path": "layers.19.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 138, "layer_count": 196, "elapsed_seconds": 2.8368008959999997, "weighted_error_rtn": 1860169.537109375, "weighted_error_gptq_base_scale": 1209434.3642578125, "weighted_error_calibrated": 945948.8876953125, "improvement_vs_rtn": 0.4914716810354409 }, { "path": "layers.19.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 139, "layer_count": 196, "elapsed_seconds": 2.8461509370000044, "weighted_error_rtn": 1245240.375, "weighted_error_gptq_base_scale": 948453.185546875, "weighted_error_calibrated": 815202.2412109375, "improvement_vs_rtn": 0.3453454790116828 }, { "path": "layers.19.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 140, "layer_count": 196, "elapsed_seconds": 0.5032528969999817, "weighted_error_rtn": 118086.0107421875, "weighted_error_gptq_base_scale": 103833.662109375, "weighted_error_calibrated": 82739.3427734375, "improvement_vs_rtn": 0.2993298507299139 }, { "path": "layers.20.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 141, "layer_count": 196, "elapsed_seconds": 0.4840516570000091, "weighted_error_rtn": 418852.4765625, "weighted_error_gptq_base_scale": 292541.490234375, "weighted_error_calibrated": 248662.671875, "improvement_vs_rtn": 0.406323978514437 }, { "path": "layers.20.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 142, "layer_count": 196, "elapsed_seconds": 0.0809733309999956, "weighted_error_rtn": 149956.15625, "weighted_error_gptq_base_scale": 96978.78515625, "weighted_error_calibrated": 65169.18359375, "improvement_vs_rtn": 0.5654117495176194 }, { "path": "layers.20.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 143, "layer_count": 196, "elapsed_seconds": 0.08076636599997755, "weighted_error_rtn": 100261.54296875, "weighted_error_gptq_base_scale": 77818.7109375, "weighted_error_calibrated": 65705.40234375, "improvement_vs_rtn": 0.34465997232628487 }, { "path": "layers.20.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 144, "layer_count": 196, "elapsed_seconds": 0.4938739599999735, "weighted_error_rtn": 160459.11328125, "weighted_error_gptq_base_scale": 114244.7412109375, "weighted_error_calibrated": 95886.8857421875, "improvement_vs_rtn": 0.40242168997831496 }, { "path": "layers.20.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 145, "layer_count": 196, "elapsed_seconds": 2.8256271729999867, "weighted_error_rtn": 1589631.625, "weighted_error_gptq_base_scale": 1019500.2109375, "weighted_error_calibrated": 806513.3818359375, "improvement_vs_rtn": 0.49264133328000603 }, { "path": "layers.20.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 146, "layer_count": 196, "elapsed_seconds": 2.8438213469999596, "weighted_error_rtn": 1077928.0263671875, "weighted_error_gptq_base_scale": 779445.93359375, "weighted_error_calibrated": 672991.50390625, "improvement_vs_rtn": 0.37566192969826273 }, { "path": "layers.20.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 147, "layer_count": 196, "elapsed_seconds": 0.5037482320000208, "weighted_error_rtn": 70475.3388671875, "weighted_error_gptq_base_scale": 58812.21142578125, "weighted_error_calibrated": 48641.6435546875, "improvement_vs_rtn": 0.3098061770748791 }, { "path": "layers.21.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 148, "layer_count": 196, "elapsed_seconds": 0.48150082000000793, "weighted_error_rtn": 379890.525390625, "weighted_error_gptq_base_scale": 260566.48046875, "weighted_error_calibrated": 213614.703125, "improvement_vs_rtn": 0.43769404908072074 }, { "path": "layers.21.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 149, "layer_count": 196, "elapsed_seconds": 0.08097409999999172, "weighted_error_rtn": 105641.9296875, "weighted_error_gptq_base_scale": 69707.86328125, "weighted_error_calibrated": 49585.09765625, "improvement_vs_rtn": 0.5306305195017929 }, { "path": "layers.21.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 150, "layer_count": 196, "elapsed_seconds": 0.08084807099999125, "weighted_error_rtn": 84378.4140625, "weighted_error_gptq_base_scale": 62007.064453125, "weighted_error_calibrated": 51980.81640625, "improvement_vs_rtn": 0.38395599178070294 }, { "path": "layers.21.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 151, "layer_count": 196, "elapsed_seconds": 0.49059815999999046, "weighted_error_rtn": 100670.029296875, "weighted_error_gptq_base_scale": 51339.20263671875, "weighted_error_calibrated": 41424.329345703125, "improvement_vs_rtn": 0.5885137847378275 }, { "path": "layers.21.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 152, "layer_count": 196, "elapsed_seconds": 2.8314578370000163, "weighted_error_rtn": 2372306.228515625, "weighted_error_gptq_base_scale": 1367573.296875, "weighted_error_calibrated": 1068581.404296875, "improvement_vs_rtn": 0.549560089902265 }, { "path": "layers.21.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 153, "layer_count": 196, "elapsed_seconds": 2.829801846999999, "weighted_error_rtn": 1594216.798828125, "weighted_error_gptq_base_scale": 1039103.5126953125, "weighted_error_calibrated": 887652.9482421875, "improvement_vs_rtn": 0.44320436913305494 }, { "path": "layers.21.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 154, "layer_count": 196, "elapsed_seconds": 0.5038280480000026, "weighted_error_rtn": 165940.4814453125, "weighted_error_gptq_base_scale": 134326.1982421875, "weighted_error_calibrated": 106250.650390625, "improvement_vs_rtn": 0.3597062665770253 }, { "path": "layers.22.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 155, "layer_count": 196, "elapsed_seconds": 0.4830946160000167, "weighted_error_rtn": 396350.62890625, "weighted_error_gptq_base_scale": 260835.71875, "weighted_error_calibrated": 213982.9892578125, "improvement_vs_rtn": 0.4601169427980736 }, { "path": "layers.22.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 156, "layer_count": 196, "elapsed_seconds": 0.08151081200003318, "weighted_error_rtn": 113962.4609375, "weighted_error_gptq_base_scale": 70879.154296875, "weighted_error_calibrated": 51419.337890625, "improvement_vs_rtn": 0.5488046022556085 }, { "path": "layers.22.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 157, "layer_count": 196, "elapsed_seconds": 0.08195592900000293, "weighted_error_rtn": 108407.1328125, "weighted_error_gptq_base_scale": 87741.966796875, "weighted_error_calibrated": 63956.8359375, "improvement_vs_rtn": 0.41003110885591665 }, { "path": "layers.22.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 158, "layer_count": 196, "elapsed_seconds": 0.49298171899999943, "weighted_error_rtn": 102539.490234375, "weighted_error_gptq_base_scale": 49979.05517578125, "weighted_error_calibrated": 41834.739013671875, "improvement_vs_rtn": 0.59201338998224 }, { "path": "layers.22.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 159, "layer_count": 196, "elapsed_seconds": 2.8770035300000245, "weighted_error_rtn": 2609057.4765625, "weighted_error_gptq_base_scale": 1467781.515625, "weighted_error_calibrated": 1140194.23828125, "improvement_vs_rtn": 0.5629861555279015 }, { "path": "layers.22.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 160, "layer_count": 196, "elapsed_seconds": 2.8425175059999788, "weighted_error_rtn": 1730508.98046875, "weighted_error_gptq_base_scale": 1079489.71875, "weighted_error_calibrated": 917256.6650390625, "improvement_vs_rtn": 0.469949780445172 }, { "path": "layers.22.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 161, "layer_count": 196, "elapsed_seconds": 0.5060013920000301, "weighted_error_rtn": 189622.0107421875, "weighted_error_gptq_base_scale": 138972.705078125, "weighted_error_calibrated": 114813.234375, "improvement_vs_rtn": 0.3945152573500471 }, { "path": "layers.23.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 162, "layer_count": 196, "elapsed_seconds": 0.48132592199999635, "weighted_error_rtn": 424631.3515625, "weighted_error_gptq_base_scale": 263844.138671875, "weighted_error_calibrated": 224039.7568359375, "improvement_vs_rtn": 0.4723899777735515 }, { "path": "layers.23.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 163, "layer_count": 196, "elapsed_seconds": 0.08092179799996302, "weighted_error_rtn": 129524.40625, "weighted_error_gptq_base_scale": 68689.74609375, "weighted_error_calibrated": 49050.044921875, "improvement_vs_rtn": 0.6213065449055089 }, { "path": "layers.23.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 164, "layer_count": 196, "elapsed_seconds": 0.0831657269999937, "weighted_error_rtn": 163940.8671875, "weighted_error_gptq_base_scale": 133574.33984375, "weighted_error_calibrated": 92909.736328125, "improvement_vs_rtn": 0.4332728750186269 }, { "path": "layers.23.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 165, "layer_count": 196, "elapsed_seconds": 0.4955932759999655, "weighted_error_rtn": 101621.2392578125, "weighted_error_gptq_base_scale": 45366.630126953125, "weighted_error_calibrated": 37511.4345703125, "improvement_vs_rtn": 0.6308701326191644 }, { "path": "layers.23.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 166, "layer_count": 196, "elapsed_seconds": 2.8548080560000244, "weighted_error_rtn": 2440427.29296875, "weighted_error_gptq_base_scale": 1379387.400390625, "weighted_error_calibrated": 1089488.064453125, "improvement_vs_rtn": 0.5535666776092412 }, { "path": "layers.23.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 167, "layer_count": 196, "elapsed_seconds": 2.844041377999986, "weighted_error_rtn": 1791292.11328125, "weighted_error_gptq_base_scale": 1097123.087890625, "weighted_error_calibrated": 931624.578125, "improvement_vs_rtn": 0.4799147658734061 }, { "path": "layers.23.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 168, "layer_count": 196, "elapsed_seconds": 0.5105992660000425, "weighted_error_rtn": 241202.259765625, "weighted_error_gptq_base_scale": 172269.318359375, "weighted_error_calibrated": 138178.9658203125, "improvement_vs_rtn": 0.42712408269068336 }, { "path": "layers.24.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 169, "layer_count": 196, "elapsed_seconds": 0.4820754319999878, "weighted_error_rtn": 375393.0234375, "weighted_error_gptq_base_scale": 228282.1953125, "weighted_error_calibrated": 190278.107421875, "improvement_vs_rtn": 0.4931229523674011 }, { "path": "layers.24.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 170, "layer_count": 196, "elapsed_seconds": 0.08897465799998372, "weighted_error_rtn": 92980.734375, "weighted_error_gptq_base_scale": 56592.20703125, "weighted_error_calibrated": 42243.755859375, "improvement_vs_rtn": 0.5456719486748515 }, { "path": "layers.24.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 171, "layer_count": 196, "elapsed_seconds": 0.0841290299999855, "weighted_error_rtn": 247470.4375, "weighted_error_gptq_base_scale": 212305.29296875, "weighted_error_calibrated": 142235.85546875, "improvement_vs_rtn": 0.4252410231070529 }, { "path": "layers.24.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 172, "layer_count": 196, "elapsed_seconds": 0.48209221999997, "weighted_error_rtn": 235107.947265625, "weighted_error_gptq_base_scale": 133507.2314453125, "weighted_error_calibrated": 110204.06640625, "improvement_vs_rtn": 0.5312618408354295 }, { "path": "layers.24.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 173, "layer_count": 196, "elapsed_seconds": 2.8403512680000063, "weighted_error_rtn": 2178362.171875, "weighted_error_gptq_base_scale": 1277788.861328125, "weighted_error_calibrated": 1043570.650390625, "improvement_vs_rtn": 0.5209379487652488 }, { "path": "layers.24.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 174, "layer_count": 196, "elapsed_seconds": 2.8329629520000026, "weighted_error_rtn": 1877464.224609375, "weighted_error_gptq_base_scale": 1155895.15625, "weighted_error_calibrated": 987821.021484375, "improvement_vs_rtn": 0.4738536114104114 }, { "path": "layers.24.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 175, "layer_count": 196, "elapsed_seconds": 0.4975533460000179, "weighted_error_rtn": 210516.9013671875, "weighted_error_gptq_base_scale": 152426.947265625, "weighted_error_calibrated": 117839.625, "improvement_vs_rtn": 0.44023674947380154 }, { "path": "layers.25.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 176, "layer_count": 196, "elapsed_seconds": 0.49216885500004537, "weighted_error_rtn": 338976.1328125, "weighted_error_gptq_base_scale": 230533.41796875, "weighted_error_calibrated": 204969.1552734375, "improvement_vs_rtn": 0.3953286516876592 }, { "path": "layers.25.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 177, "layer_count": 196, "elapsed_seconds": 0.09044252099999994, "weighted_error_rtn": 94150.21875, "weighted_error_gptq_base_scale": 67746.09765625, "weighted_error_calibrated": 48278.4375, "improvement_vs_rtn": 0.4872190618250688 }, { "path": "layers.25.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 178, "layer_count": 196, "elapsed_seconds": 0.0809960689999798, "weighted_error_rtn": 460255.578125, "weighted_error_gptq_base_scale": 420302.71875, "weighted_error_calibrated": 243937.9453125, "improvement_vs_rtn": 0.46999459233876073 }, { "path": "layers.25.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 179, "layer_count": 196, "elapsed_seconds": 0.48120142800001986, "weighted_error_rtn": 208645.1923828125, "weighted_error_gptq_base_scale": 112633.8623046875, "weighted_error_calibrated": 95388.1337890625, "improvement_vs_rtn": 0.5428213192947731 }, { "path": "layers.25.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 180, "layer_count": 196, "elapsed_seconds": 2.8602341159999582, "weighted_error_rtn": 1932422.30859375, "weighted_error_gptq_base_scale": 1118445.7548828125, "weighted_error_calibrated": 940041.3583984375, "improvement_vs_rtn": 0.5135424828113693 }, { "path": "layers.25.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 181, "layer_count": 196, "elapsed_seconds": 2.8530012489999876, "weighted_error_rtn": 1853887.708984375, "weighted_error_gptq_base_scale": 1129628.15234375, "weighted_error_calibrated": 963102.2763671875, "improvement_vs_rtn": 0.4804958942767851 }, { "path": "layers.25.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 182, "layer_count": 196, "elapsed_seconds": 0.49377206399998386, "weighted_error_rtn": 264701.92578125, "weighted_error_gptq_base_scale": 157808.875, "weighted_error_calibrated": 126989.453125, "improvement_vs_rtn": 0.5202548951988197 }, { "path": "layers.26.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 183, "layer_count": 196, "elapsed_seconds": 0.4847801569999888, "weighted_error_rtn": 395596.578125, "weighted_error_gptq_base_scale": 252490.509765625, "weighted_error_calibrated": 192997.681640625, "improvement_vs_rtn": 0.5121351085609192 }, { "path": "layers.26.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 184, "layer_count": 196, "elapsed_seconds": 0.09146888800000852, "weighted_error_rtn": 88628.97265625, "weighted_error_gptq_base_scale": 57097.90625, "weighted_error_calibrated": 40457.03125, "improvement_vs_rtn": 0.5435236352460753 }, { "path": "layers.26.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 185, "layer_count": 196, "elapsed_seconds": 0.08138046400000576, "weighted_error_rtn": 730997.21875, "weighted_error_gptq_base_scale": 659178.625, "weighted_error_calibrated": 404142.25, "improvement_vs_rtn": 0.4471357214038648 }, { "path": "layers.26.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 186, "layer_count": 196, "elapsed_seconds": 0.4853933520000169, "weighted_error_rtn": 1239128.3828125, "weighted_error_gptq_base_scale": 709657.23046875, "weighted_error_calibrated": 561605.04296875, "improvement_vs_rtn": 0.5467741270730542 }, { "path": "layers.26.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 187, "layer_count": 196, "elapsed_seconds": 2.8558329319999984, "weighted_error_rtn": 2182786.39453125, "weighted_error_gptq_base_scale": 1279098.474609375, "weighted_error_calibrated": 1087560.4384765625, "improvement_vs_rtn": 0.5017559019053193 }, { "path": "layers.26.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 188, "layer_count": 196, "elapsed_seconds": 2.8492842080000287, "weighted_error_rtn": 2305980.880859375, "weighted_error_gptq_base_scale": 1393965.4765625, "weighted_error_calibrated": 1175449.6005859375, "improvement_vs_rtn": 0.49026047425515507 }, { "path": "layers.26.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 189, "layer_count": 196, "elapsed_seconds": 0.5056334160000233, "weighted_error_rtn": 27475175.46875, "weighted_error_gptq_base_scale": 27180503.28125, "weighted_error_calibrated": 14930181.1875, "improvement_vs_rtn": 0.4565937821040872 }, { "path": "layers.27.self_attn.q_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 190, "layer_count": 196, "elapsed_seconds": 0.48756065000003446, "weighted_error_rtn": 490870.64453125, "weighted_error_gptq_base_scale": 298722.392578125, "weighted_error_calibrated": 238933.708984375, "improvement_vs_rtn": 0.5132450643640722 }, { "path": "layers.27.self_attn.k_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 191, "layer_count": 196, "elapsed_seconds": 0.08174685000000181, "weighted_error_rtn": 116472.17578125, "weighted_error_gptq_base_scale": 69894.94921875, "weighted_error_calibrated": 48393.53515625, "improvement_vs_rtn": 0.5845056140520686 }, { "path": "layers.27.self_attn.v_proj", "shape": [ 256, 1536 ], "calibration_tokens": 32768, "layer_index": 192, "layer_count": 196, "elapsed_seconds": 0.09366602599999396, "weighted_error_rtn": 893903.625, "weighted_error_gptq_base_scale": 752757.8125, "weighted_error_calibrated": 459714.9375, "improvement_vs_rtn": 0.4857220346320891 }, { "path": "layers.27.self_attn.o_proj", "shape": [ 1536, 1536 ], "calibration_tokens": 32768, "layer_index": 193, "layer_count": 196, "elapsed_seconds": 0.484694764999972, "weighted_error_rtn": 2373026.65625, "weighted_error_gptq_base_scale": 1356621.703125, "weighted_error_calibrated": 1120156.140625, "improvement_vs_rtn": 0.5279631024477667 }, { "path": "layers.27.mlp.gate_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 194, "layer_count": 196, "elapsed_seconds": 2.858110773999954, "weighted_error_rtn": 4333538.046875, "weighted_error_gptq_base_scale": 2031112.958984375, "weighted_error_calibrated": 1703743.900390625, "improvement_vs_rtn": 0.606846903855101 }, { "path": "layers.27.mlp.up_proj", "shape": [ 8960, 1536 ], "calibration_tokens": 32768, "layer_index": 195, "layer_count": 196, "elapsed_seconds": 2.845924155999967, "weighted_error_rtn": 5362745.44921875, "weighted_error_gptq_base_scale": 2045888.876953125, "weighted_error_calibrated": 1696763.61328125, "improvement_vs_rtn": 0.6836016869813509 }, { "path": "layers.27.mlp.down_proj", "shape": [ 1536, 8960 ], "calibration_tokens": 32768, "layer_index": 196, "layer_count": 196, "elapsed_seconds": 0.5081464849999975, "weighted_error_rtn": 1261439.03125, "weighted_error_gptq_base_scale": 554604.138671875, "weighted_error_calibrated": 399711.78515625, "improvement_vs_rtn": 0.6831303176340098 } ], "improvement_vs_rtn": 0.4247495674721419 }