| { |
| "model_id": "letxbe/qwen2-7b-BoundingDocs-rephrased", |
| "base_model_id": "letxbe/qwen2-7b-BoundingDocs-rephrased", |
| "processor_id": "Qwen/Qwen2-VL-7B-Instruct", |
| "dataset_id": "letxbe/BoundingDocs", |
| "method": "static_w8a8_smoothquant_style", |
| "format_version": 2, |
| "quant_module_class": "W8A8SmoothLinear", |
| "smoothquant_alpha": 0.5, |
| "converted_layers": 327, |
| "skipped_layers": 0, |
| "quant_meta": { |
| "model.visual.blocks.0.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.013058592015364039, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.0.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.005015721001962977, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.0.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.022689226105457216, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.0.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.01226213504010298, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.1.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.01299497461694432, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.1.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.008844585869255968, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.1.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01934741050239623, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.1.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.006352477655636044, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.2.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.0174557355445201, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.2.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.0069069745033744755, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.2.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.010665267471253402, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.2.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.009205215559231015, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.3.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.02207961044912263, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.3.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.006421548644388754, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.3.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.010995670566408653, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.3.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.009494174183822993, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.4.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.015408067252692274, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.4.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.00744820157373984, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.4.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.012565937567883588, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.4.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.008713548577676608, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.5.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.010462606047082134, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.5.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.007633128504114827, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.5.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.011509261732026347, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.5.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.007014970610460897, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.6.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.011709320263599787, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.6.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.005283016858138438, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.6.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01479891810830184, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.6.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.00873193309063048, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.7.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.013217586231982614, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.7.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.005463831537351834, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.7.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.013557838642691064, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.7.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.007997047244094488, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.8.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.022513083585603968, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.8.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.00962225095493587, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.8.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.016712596097330407, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.8.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.009133379290423056, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.9.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014724607542743833, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.9.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.006425451575301763, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.9.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.0164236684483806, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.9.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.01196374386314332, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.10.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.012820536696066068, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.10.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.006016723283632534, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.10.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.0169328160173311, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.10.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.011545721001512422, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.11.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014503340082844412, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.11.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.006066088601360171, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.11.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.016097098823607436, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.11.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.015256804744089683, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.12.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.015001395555931753, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.12.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.006537278806130717, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.12.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.016522905019324595, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.12.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.017558677928654228, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.13.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.015230894088745117, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.13.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.005747074217308225, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.13.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.013493245042215182, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.13.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.019083023071289062, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.14.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.01589116712254802, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.14.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.00570369940104447, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.14.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01756293567146842, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.14.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.02456905334953248, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.15.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014987766273378387, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.15.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.007739594132881465, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.15.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.017646231989222248, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.15.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.01963258540536475, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.16.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014867702807028463, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.16.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.005611331444086991, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.16.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.017360142835481897, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.16.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.01572891764753447, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.17.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.012322933655085527, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.17.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.0068853689929631754, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.17.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.021783440131840744, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.17.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.02555163075604777, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.18.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014477257653484194, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.18.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.009281032667385312, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.18.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01921030480091966, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.18.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.04086608961811216, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.19.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.015321775684206505, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.19.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.0066865748307836335, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.19.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01889635071041077, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.19.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.017186799387293537, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.20.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014851434962955986, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.20.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.005967403960040235, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.20.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01988548368919553, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.20.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.013951839424493745, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.21.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.013912627077478123, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.21.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.011704976164449857, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.21.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01975512880039966, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.21.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.02030630186786802, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.22.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014595573342691256, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.22.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.007727973573789822, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.22.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.021446421390443338, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.22.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.019704251777468703, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.23.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.01592456074211541, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.23.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.008090548627958523, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.23.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.023214841452170544, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.23.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.026059163836982306, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.24.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.017067792847400575, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.24.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.007447630870999314, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.24.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.031022606872198148, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.24.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.028092566437608615, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.25.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.01583310562794603, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.25.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.009170232795354888, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.25.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.028548377705371285, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.25.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.018723814506230393, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.26.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.015311120063301147, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.26.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.008967384578674797, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.26.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.03503482548270639, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.26.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.02070761290122205, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.27.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.017509310264287033, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.27.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.00825775889899787, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.27.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.027688989489097295, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.27.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.022381632346806565, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.28.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.019842191005316305, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.28.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.008579958142258051, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.28.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.034220177357591045, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.28.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.026569037925539992, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.29.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.019085998610248716, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.29.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.010472141851590374, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.29.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.03460992227389118, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.29.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.044394939903199204, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.30.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.01766423353060024, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.30.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.01588429240729865, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.30.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.028332124544879583, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.30.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.03656830750112459, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.31.attn.qkv": { |
| "in_features": 1280, |
| "out_features": 3840, |
| "act_scale": 0.014738044400853434, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.31.attn.proj": { |
| "in_features": 1280, |
| "out_features": 1280, |
| "act_scale": 0.020153285950187623, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.31.mlp.fc1": { |
| "in_features": 1280, |
| "out_features": 5120, |
| "act_scale": 0.01568901820445624, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.blocks.31.mlp.fc2": { |
| "in_features": 5120, |
| "out_features": 1280, |
| "act_scale": 0.1372537688007505, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.merger.mlp.0": { |
| "in_features": 5120, |
| "out_features": 5120, |
| "act_scale": 0.008264830731970119, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.visual.merger.mlp.2": { |
| "in_features": 5120, |
| "out_features": 3584, |
| "act_scale": 0.008555569986658773, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.01953330753356453, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01529226340646819, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.008840601275286337, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.005158939699488362, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.020522117614746094, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.00804166906461941, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.0.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.00946472104140154, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.019910016397791584, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.015997169524665893, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.004661967904548946, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.003689274309188362, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.024118410320732538, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.028101992419385536, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.1.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.015522642398443748, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.013075172431825653, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01332774594074159, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.006630623434472272, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.0047334662572605405, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.030007138965636726, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.024669117814912572, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.2.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.02236598870885654, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.012795274651895358, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.012875614203805999, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.005618176122350017, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.004145878975785624, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.027995376136359267, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.03481524760328879, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.3.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.0941666130005844, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.016053254210104152, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.014967681854728639, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.009513013944851132, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.004991027783221147, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.022710770133912095, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.02253127098083496, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.4.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.05879103292630413, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.014688874792864941, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.014583515369985986, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.009067288533909115, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.006945680445573461, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.022198791578998716, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.0347787676833746, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.5.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.05613904111967312, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.015444005568196454, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.012391416106637068, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01119628853685274, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.006362457913676585, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.024242472460889442, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.022583905167467014, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.6.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.02453051777336541, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.017803229684904803, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.014832822356637068, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.009144973567151648, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.006266298725849061, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.018840294184647208, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.015237881442693274, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.7.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.031839374482162355, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.017244939728984682, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01515336487236924, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.00930744644225113, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.007323991595290777, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.016757056469053733, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.018776921775397352, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.8.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.029563291804997, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.01954604133846253, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.015926670840406044, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.006540489009046179, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.010204918741241215, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01994704449270654, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.018006469321063184, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.9.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.04155524321428434, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.018466716676246463, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.015691882043372928, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.008046697443864477, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.007237093185815286, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01791983514320193, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.016117065910279282, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.10.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.015522642398443748, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.015953465709536093, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.019926163155262865, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.009570753480505756, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.008202496476060762, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.014006420383303185, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.022370935425044983, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.11.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.03829344051090751, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.015434507309921144, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.018373036947776013, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.010324755052881917, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.008430421821714386, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01213280328615444, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.021014630325197233, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.12.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.022604925425972526, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.019029159245528574, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.019423982289832407, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.0082286080037515, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.008317971792746716, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01599853057560958, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.014360534863209162, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.13.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.018865388209425557, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.014979337144085742, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.02217640651492622, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.012561192662697139, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.011742702619297298, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.012810755902387965, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01946902838278943, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.14.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.01819218237569013, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.018790770703413354, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.018373035070464366, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01222852646835207, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.006755958861253393, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01222131195969469, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.013427036015067513, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.15.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.016127836047195074, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.01493369969796008, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.02071496633094127, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.011933303254795826, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.008233521866986133, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01605163033552996, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01970166108739658, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.16.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.0110542351805319, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.014674762102562612, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.018540902400579977, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.010319874042601098, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.010709439675638996, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.010568765204722486, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.018275405478289746, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.17.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.01770029668732891, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.019733183027252438, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.020599624303382212, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.010792902135473537, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.007693651154285341, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.01423262235686535, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.011693454164219654, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.18.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.025894405334953248, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.020787180878045992, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.02119884903975359, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.011487232418510858, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.009082626170060766, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.011785403950007881, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.011364712489871528, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.19.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.024989710079403375, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.013816620421221875, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01961098318024883, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.010185199459706704, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.00840593604590949, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.013985906998942218, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.0154962314395454, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.20.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.024939266715462753, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.0167091981632503, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.02066663681991457, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.011174764220170148, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.008848678408645269, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.015170712170638437, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.013777329227117103, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.21.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.03435295210109921, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.013921156642943855, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.019353718269528367, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.012059989876634493, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.007871056635548749, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.014820156134958342, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.012657005017197976, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.22.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.029494860040859914, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.018486276386291023, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.017934134626013086, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01271718224202554, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.009331758566728727, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.017162784816711907, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.014909640071898933, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.23.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.04040793546541469, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.01624761228486309, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.019860990404144047, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.013822918801795779, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.011292119664470041, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.018050443469070074, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.0191757528800664, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.24.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.05960619168018731, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.015364412247665285, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.017897806768342267, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.01793165657463975, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.009667346796651525, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.0283089078317477, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.03614818017313799, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.25.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.07223321509173536, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.017873474932092383, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.019200206741573304, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.020494994216077908, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.012540879212026521, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.03283811929657703, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.05497732312660518, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.26.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.41332022599347934, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.self_attn.q_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.04454606724536325, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.self_attn.k_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.021689950011846586, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.self_attn.v_proj": { |
| "in_features": 3584, |
| "out_features": 512, |
| "act_scale": 0.05705929928877222, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.self_attn.o_proj": { |
| "in_features": 3584, |
| "out_features": 3584, |
| "act_scale": 0.042185993645134874, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.mlp.gate_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.028611301437137634, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.mlp.up_proj": { |
| "in_features": 3584, |
| "out_features": 18944, |
| "act_scale": 0.034482910877137674, |
| "smoothquant_alpha": 0.5 |
| }, |
| "model.language_model.layers.27.mlp.down_proj": { |
| "in_features": 18944, |
| "out_features": 3584, |
| "act_scale": 0.0871090776338352, |
| "smoothquant_alpha": 0.5 |
| }, |
| "lm_head": { |
| "in_features": 3584, |
| "out_features": 152064, |
| "act_scale": 0.018600555855458178, |
| "smoothquant_alpha": 0.5 |
| } |
| }, |
| "self_contained_bundle": true, |
| "requires_base_model_download": false |
| } |
