numind/NuExtract3-FP8
513k
1{2 "architectures": [3 "Qwen3_5ForConditionalGeneration"4 ],5 "bos_token_id": null,6 "dtype": "bfloat16",7 "eos_token_id": 248046,8 "head_dim": 256,9 "hidden_size": 2560,10 "image_token_id": 248056,11 "model_type": "qwen3_5",12 "num_attention_heads": 16,13 "num_hidden_layers": 32,14 "num_key_value_heads": 4,15 "pad_token_id": 248044,16 "quantization_config": {17 "config_groups": {18 "group_0": {19 "format": "float-quantized",20 "input_activations": {21 "actorder": null,22 "block_structure": null,23 "dynamic": false,24 "group_size": null,25 "num_bits": 8,26 "observer": "memoryless_minmax",27 "observer_kwargs": {},28 "scale_dtype": null,29 "strategy": "tensor",30 "symmetric": true,31 "type": "float",32 "zp_dtype": null33 },34 "output_activations": null,35 "targets": [36 "Linear"37 ],38 "weights": {39 "actorder": null,40 "block_structure": null,41 "dynamic": false,42 "group_size": null,43 "num_bits": 8,44 "observer": "memoryless_minmax",45 "observer_kwargs": {},46 "scale_dtype": null,47 "strategy": "tensor",48 "symmetric": true,49 "type": "float",50 "zp_dtype": null51 }52 }53 },54 "format": "float-quantized",55 "global_compression_ratio": null,56 "ignore": [57 "model.visual.blocks.0.attn.qkv",58 "model.visual.blocks.0.attn.proj",59 "model.visual.blocks.0.mlp.linear_fc1",60 "model.visual.blocks.0.mlp.linear_fc2",61 "model.visual.blocks.1.attn.qkv",62 "model.visual.blocks.1.attn.proj",63 "model.visual.blocks.1.mlp.linear_fc1",64 "model.visual.blocks.1.mlp.linear_fc2",65 "model.visual.blocks.2.attn.qkv",66 "model.visual.blocks.2.attn.proj",67 "model.visual.blocks.2.mlp.linear_fc1",68 "model.visual.blocks.2.mlp.linear_fc2",69 "model.visual.blocks.3.attn.qkv",70 "model.visual.blocks.3.attn.proj",71 "model.visual.blocks.3.mlp.linear_fc1",72 "model.visual.blocks.3.mlp.linear_fc2",73 "model.visual.blocks.4.attn.qkv",74 "model.visual.blocks.4.attn.proj",75 "model.visual.blocks.4.mlp.linear_fc1",76 "model.visual.blocks.4.mlp.linear_fc2",77 "model.visual.blocks.5.attn.qkv",78 "model.visual.blocks.5.attn.proj",79 "model.visual.blocks.5.mlp.linear_fc1",80 "model.visual.blocks.5.mlp.linear_fc2",81 "model.visual.blocks.6.attn.qkv",82 "model.visual.blocks.6.attn.proj",83 "model.visual.blocks.6.mlp.linear_fc1",84 "model.visual.blocks.6.mlp.linear_fc2",85 "model.visual.blocks.7.attn.qkv",86 "model.visual.blocks.7.attn.proj",87 "model.visual.blocks.7.mlp.linear_fc1",88 "model.visual.blocks.7.mlp.linear_fc2",89 "model.visual.blocks.8.attn.qkv",90 "model.visual.blocks.8.attn.proj",91 "model.visual.blocks.8.mlp.linear_fc1",92 "model.visual.blocks.8.mlp.linear_fc2",93 "model.visual.blocks.9.attn.qkv",94 "model.visual.blocks.9.attn.proj",95 "model.visual.blocks.9.mlp.linear_fc1",96 "model.visual.blocks.9.mlp.linear_fc2",97 "model.visual.blocks.10.attn.qkv",98 "model.visual.blocks.10.attn.proj",99 "model.visual.blocks.10.mlp.linear_fc1",100 "model.visual.blocks.10.mlp.linear_fc2",101 "model.visual.blocks.11.attn.qkv",102 "model.visual.blocks.11.attn.proj",103 "model.visual.blocks.11.mlp.linear_fc1",104 "model.visual.blocks.11.mlp.linear_fc2",105 "model.visual.blocks.12.attn.qkv",106 "model.visual.blocks.12.attn.proj",107 "model.visual.blocks.12.mlp.linear_fc1",108 "model.visual.blocks.12.mlp.linear_fc2",109 "model.visual.blocks.13.attn.qkv",110 "model.visual.blocks.13.attn.proj",111 "model.visual.blocks.13.mlp.linear_fc1",112 "model.visual.blocks.13.mlp.linear_fc2",113 "model.visual.blocks.14.attn.qkv",114 "model.visual.blocks.14.attn.proj",115 "model.visual.blocks.14.mlp.linear_fc1",116 "model.visual.blocks.14.mlp.linear_fc2",117 "model.visual.blocks.15.attn.qkv",118 "model.visual.blocks.15.attn.proj",119 "model.visual.blocks.15.mlp.linear_fc1",120 "model.visual.blocks.15.mlp.linear_fc2",121 "model.visual.blocks.16.attn.qkv",122 "model.visual.blocks.16.attn.proj",123 "model.visual.blocks.16.mlp.linear_fc1",124 "model.visual.blocks.16.mlp.linear_fc2",125 "model.visual.blocks.17.attn.qkv",126 "model.visual.blocks.17.attn.proj",127 "model.visual.blocks.17.mlp.linear_fc1",128 "model.visual.blocks.17.mlp.linear_fc2",129 "model.visual.blocks.18.attn.qkv",130 "model.visual.blocks.18.attn.proj",131 "model.visual.blocks.18.mlp.linear_fc1",132 "model.visual.blocks.18.mlp.linear_fc2",133 "model.visual.blocks.19.attn.qkv",134 "model.visual.blocks.19.attn.proj",135 "model.visual.blocks.19.mlp.linear_fc1",136 "model.visual.blocks.19.mlp.linear_fc2",137 "model.visual.blocks.20.attn.qkv",138 "model.visual.blocks.20.attn.proj",139 "model.visual.blocks.20.mlp.linear_fc1",140 "model.visual.blocks.20.mlp.linear_fc2",141 "model.visual.blocks.21.attn.qkv",142 "model.visual.blocks.21.attn.proj",143 "model.visual.blocks.21.mlp.linear_fc1",144 "model.visual.blocks.21.mlp.linear_fc2",145 "model.visual.blocks.22.attn.qkv",146 "model.visual.blocks.22.attn.proj",147 "model.visual.blocks.22.mlp.linear_fc1",148 "model.visual.blocks.22.mlp.linear_fc2",149 "model.visual.blocks.23.attn.qkv",150 "model.visual.blocks.23.attn.proj",151 "model.visual.blocks.23.mlp.linear_fc1",152 "model.visual.blocks.23.mlp.linear_fc2",153 "model.visual.merger.linear_fc1",154 "model.visual.merger.linear_fc2",155 "model.language_model.layers.0.linear_attn.out_proj",156 "model.language_model.layers.0.linear_attn.in_proj_qkv",157 "model.language_model.layers.0.linear_attn.in_proj_z",158 "model.language_model.layers.0.linear_attn.in_proj_b",159 "model.language_model.layers.0.linear_attn.in_proj_a",160 "model.language_model.layers.1.linear_attn.out_proj",161 "model.language_model.layers.1.linear_attn.in_proj_qkv",162 "model.language_model.layers.1.linear_attn.in_proj_z",163 "model.language_model.layers.1.linear_attn.in_proj_b",164 "model.language_model.layers.1.linear_attn.in_proj_a",165 "model.language_model.layers.2.linear_attn.out_proj",166 "model.language_model.layers.2.linear_attn.in_proj_qkv",167 "model.language_model.layers.2.linear_attn.in_proj_z",168 "model.language_model.layers.2.linear_attn.in_proj_b",169 "model.language_model.layers.2.linear_attn.in_proj_a",170 "model.language_model.layers.4.linear_attn.out_proj",171 "model.language_model.layers.4.linear_attn.in_proj_qkv",172 "model.language_model.layers.4.linear_attn.in_proj_z",173 "model.language_model.layers.4.linear_attn.in_proj_b",174 "model.language_model.layers.4.linear_attn.in_proj_a",175 "model.language_model.layers.5.linear_attn.out_proj",176 "model.language_model.layers.5.linear_attn.in_proj_qkv",177 "model.language_model.layers.5.linear_attn.in_proj_z",178 "model.language_model.layers.5.linear_attn.in_proj_b",179 "model.language_model.layers.5.linear_attn.in_proj_a",180 "model.language_model.layers.6.linear_attn.out_proj",181 "model.language_model.layers.6.linear_attn.in_proj_qkv",182 "model.language_model.layers.6.linear_attn.in_proj_z",183 "model.language_model.layers.6.linear_attn.in_proj_b",184 "model.language_model.layers.6.linear_attn.in_proj_a",185 "model.language_model.layers.8.linear_attn.out_proj",186 "model.language_model.layers.8.linear_attn.in_proj_qkv",187 "model.language_model.layers.8.linear_attn.in_proj_z",188 "model.language_model.layers.8.linear_attn.in_proj_b",189 "model.language_model.layers.8.linear_attn.in_proj_a",190 "model.language_model.layers.9.linear_attn.out_proj",191 "model.language_model.layers.9.linear_attn.in_proj_qkv",192 "model.language_model.layers.9.linear_attn.in_proj_z",193 "model.language_model.layers.9.linear_attn.in_proj_b",194 "model.language_model.layers.9.linear_attn.in_proj_a",195 "model.language_model.layers.10.linear_attn.out_proj",196 "model.language_model.layers.10.linear_attn.in_proj_qkv",197 "model.language_model.layers.10.linear_attn.in_proj_z",198 "model.language_model.layers.10.linear_attn.in_proj_b",199 "model.language_model.layers.10.linear_attn.in_proj_a",200 "model.language_model.layers.12.linear_attn.out_proj",201 "model.language_model.layers.12.linear_attn.in_proj_qkv",202 "model.language_model.layers.12.linear_attn.in_proj_z",203 "model.language_model.layers.12.linear_attn.in_proj_b",204 "model.language_model.layers.12.linear_attn.in_proj_a",205 "model.language_model.layers.13.linear_attn.out_proj",206 "model.language_model.layers.13.linear_attn.in_proj_qkv",207 "model.language_model.layers.13.linear_attn.in_proj_z",208 "model.language_model.layers.13.linear_attn.in_proj_b",209 "model.language_model.layers.13.linear_attn.in_proj_a",210 "model.language_model.layers.14.linear_attn.out_proj",211 "model.language_model.layers.14.linear_attn.in_proj_qkv",212 "model.language_model.layers.14.linear_attn.in_proj_z",213 "model.language_model.layers.14.linear_attn.in_proj_b",214 "model.language_model.layers.14.linear_attn.in_proj_a",215 "model.language_model.layers.16.linear_attn.out_proj",216 "model.language_model.layers.16.linear_attn.in_proj_qkv",217 "model.language_model.layers.16.linear_attn.in_proj_z",218 "model.language_model.layers.16.linear_attn.in_proj_b",219 "model.language_model.layers.16.linear_attn.in_proj_a",220 "model.language_model.layers.17.linear_attn.out_proj",221 "model.language_model.layers.17.linear_attn.in_proj_qkv",222 "model.language_model.layers.17.linear_attn.in_proj_z",223 "model.language_model.layers.17.linear_attn.in_proj_b",224 "model.language_model.layers.17.linear_attn.in_proj_a",225 "model.language_model.layers.18.linear_attn.out_proj",226 "model.language_model.layers.18.linear_attn.in_proj_qkv",227 "model.language_model.layers.18.linear_attn.in_proj_z",228 "model.language_model.layers.18.linear_attn.in_proj_b",229 "model.language_model.layers.18.linear_attn.in_proj_a",230 "model.language_model.layers.20.linear_attn.out_proj",231 "model.language_model.layers.20.linear_attn.in_proj_qkv",232 "model.language_model.layers.20.linear_attn.in_proj_z",233 "model.language_model.layers.20.linear_attn.in_proj_b",234 "model.language_model.layers.20.linear_attn.in_proj_a",235 "model.language_model.layers.21.linear_attn.out_proj",236 "model.language_model.layers.21.linear_attn.in_proj_qkv",237 "model.language_model.layers.21.linear_attn.in_proj_z",238 "model.language_model.layers.21.linear_attn.in_proj_b",239 "model.language_model.layers.21.linear_attn.in_proj_a",240 "model.language_model.layers.22.linear_attn.out_proj",241 "model.language_model.layers.22.linear_attn.in_proj_qkv",242 "model.language_model.layers.22.linear_attn.in_proj_z",243 "model.language_model.layers.22.linear_attn.in_proj_b",244 "model.language_model.layers.22.linear_attn.in_proj_a",245 "model.language_model.layers.24.linear_attn.out_proj",246 "model.language_model.layers.24.linear_attn.in_proj_qkv",247 "model.language_model.layers.24.linear_attn.in_proj_z",248 "model.language_model.layers.24.linear_attn.in_proj_b",249 "model.language_model.layers.24.linear_attn.in_proj_a",250 "model.language_model.layers.25.linear_attn.out_proj",251 "model.language_model.layers.25.linear_attn.in_proj_qkv",252 "model.language_model.layers.25.linear_attn.in_proj_z",253 "model.language_model.layers.25.linear_attn.in_proj_b",254 "model.language_model.layers.25.linear_attn.in_proj_a",255 "model.language_model.layers.26.linear_attn.out_proj",256 "model.language_model.layers.26.linear_attn.in_proj_qkv",257 "model.language_model.layers.26.linear_attn.in_proj_z",258 "model.language_model.layers.26.linear_attn.in_proj_b",259 "model.language_model.layers.26.linear_attn.in_proj_a",260 "model.language_model.layers.28.linear_attn.out_proj",261 "model.language_model.layers.28.linear_attn.in_proj_qkv",262 "model.language_model.layers.28.linear_attn.in_proj_z",263 "model.language_model.layers.28.linear_attn.in_proj_b",264 "model.language_model.layers.28.linear_attn.in_proj_a",265 "model.language_model.layers.29.linear_attn.out_proj",266 "model.language_model.layers.29.linear_attn.in_proj_qkv",267 "model.language_model.layers.29.linear_attn.in_proj_z",268 "model.language_model.layers.29.linear_attn.in_proj_b",269 "model.language_model.layers.29.linear_attn.in_proj_a",270 "model.language_model.layers.30.linear_attn.out_proj",271 "model.language_model.layers.30.linear_attn.in_proj_qkv",272 "model.language_model.layers.30.linear_attn.in_proj_z",273 "model.language_model.layers.30.linear_attn.in_proj_b",274 "model.language_model.layers.30.linear_attn.in_proj_a",275 "lm_head",276 "re:^mtp.*"277 ],278 "kv_cache_scheme": {279 "actorder": null,280 "block_structure": null,281 "dynamic": false,282 "group_size": null,283 "num_bits": 8,284 "observer": "memoryless_minmax",285 "observer_kwargs": {},286 "scale_dtype": null,287 "strategy": "tensor",288 "symmetric": true,289 "type": "float",290 "zp_dtype": null291 },292 "quant_method": "compressed-tensors",293 "quantization_status": "compressed",294 "sparsity_config": {},295 "transform_config": {},296 "version": "0.15.1.a20260503"297 },298 "text_config": {299 "attention_bias": false,300 "attention_dropout": 0.0,301 "attn_output_gate": true,302 "bos_token_id": null,303 "dtype": "bfloat16",304 "eos_token_id": 248044,305 "full_attention_interval": 4,306 "head_dim": 256,307 "hidden_act": "silu",308 "hidden_size": 2560,309 "initializer_range": 0.02,310 "intermediate_size": 9216,311 "layer_types": [312 "linear_attention",313 "linear_attention",314 "linear_attention",315 "full_attention",316 "linear_attention",317 "linear_attention",318 "linear_attention",319 "full_attention",320 "linear_attention",321 "linear_attention",322 "linear_attention",323 "full_attention",324 "linear_attention",325 "linear_attention",326 "linear_attention",327 "full_attention",328 "linear_attention",329 "linear_attention",330 "linear_attention",331 "full_attention",332 "linear_attention",333 "linear_attention",334 "linear_attention",335 "full_attention",336 "linear_attention",337 "linear_attention",338 "linear_attention",339 "full_attention",340 "linear_attention",341 "linear_attention",342 "linear_attention",343 "full_attention"344 ],345 "linear_conv_kernel_dim": 4,346 "linear_key_head_dim": 128,347 "linear_num_key_heads": 16,348 "linear_num_value_heads": 32,349 "linear_value_head_dim": 128,350 "mamba_ssm_dtype": "float32",351 "max_position_embeddings": 262144,352 "mlp_only_layers": [],353 "model_type": "qwen3_5_text",354 "mtp_num_hidden_layers": 1,355 "mtp_use_dedicated_embeddings": false,356 "num_attention_heads": 16,357 "num_hidden_layers": 32,358 "num_key_value_heads": 4,359 "pad_token_id": null,360 "partial_rotary_factor": 0.25,361 "rms_norm_eps": 1e-06,362 "rope_parameters": {363 "mrope_interleaved": true,364 "mrope_section": [365 11,366 11,367 10368 ],369 "partial_rotary_factor": 0.25,370 "rope_theta": 10000000,371 "rope_type": "default"372 },373 "tie_word_embeddings": true,374 "use_cache": true,375 "vocab_size": 248320376 },377 "tie_word_embeddings": true,378 "transformers_version": "5.3.0",379 "video_token_id": 248057,380 "vision_config": {381 "deepstack_visual_indexes": [],382 "depth": 24,383 "dtype": "bfloat16",384 "hidden_act": "gelu_pytorch_tanh",385 "hidden_size": 1024,386 "in_channels": 3,387 "initializer_range": 0.02,388 "intermediate_size": 4096,389 "model_type": "qwen3_5",390 "num_heads": 16,391 "num_position_embeddings": 2304,392 "out_hidden_size": 2560,393 "patch_size": 16,394 "spatial_merge_size": 2,395 "temporal_patch_size": 2396 },397 "vision_end_token_id": 248054,398 "vision_start_token_id": 248053399}