config.json
| 1 | { |
| 2 | "_commit_hash": null, |
| 3 | "architectures": [ |
| 4 | "DPTForDepthEstimation" |
| 5 | ], |
| 6 | "attention_probs_dropout_prob": 0.0, |
| 7 | "auxiliary_loss_weight": 0.4, |
| 8 | "backbone_config": { |
| 9 | "_name_or_path": "", |
| 10 | "add_cross_attention": false, |
| 11 | "architectures": null, |
| 12 | "bad_words_ids": null, |
| 13 | "begin_suppress_tokens": null, |
| 14 | "bos_token_id": null, |
| 15 | "chunk_size_feed_forward": 0, |
| 16 | "cross_attention_hidden_size": null, |
| 17 | "decoder_start_token_id": null, |
| 18 | "depths": [ |
| 19 | 3, |
| 20 | 4, |
| 21 | 9 |
| 22 | ], |
| 23 | "diversity_penalty": 0.0, |
| 24 | "do_sample": false, |
| 25 | "drop_path_rate": 0.0, |
| 26 | "early_stopping": false, |
| 27 | "embedding_dynamic_padding": true, |
| 28 | "embedding_size": 64, |
| 29 | "encoder_no_repeat_ngram_size": 0, |
| 30 | "eos_token_id": null, |
| 31 | "exponential_decay_length_penalty": null, |
| 32 | "finetuning_task": null, |
| 33 | "forced_bos_token_id": null, |
| 34 | "forced_eos_token_id": null, |
| 35 | "global_padding": "SAME", |
| 36 | "hidden_act": "relu", |
| 37 | "hidden_sizes": [ |
| 38 | 256, |
| 39 | 512, |
| 40 | 1024, |
| 41 | 2048 |
| 42 | ], |
| 43 | "id2label": { |
| 44 | "0": "LABEL_0", |
| 45 | "1": "LABEL_1" |
| 46 | }, |
| 47 | "is_decoder": false, |
| 48 | "is_encoder_decoder": false, |
| 49 | "label2id": { |
| 50 | "LABEL_0": 0, |
| 51 | "LABEL_1": 1 |
| 52 | }, |
| 53 | "layer_type": "bottleneck", |
| 54 | "length_penalty": 1.0, |
| 55 | "max_length": 20, |
| 56 | "min_length": 0, |
| 57 | "model_type": "bit", |
| 58 | "no_repeat_ngram_size": 0, |
| 59 | "num_beam_groups": 1, |
| 60 | "num_beams": 1, |
| 61 | "num_channels": 3, |
| 62 | "num_groups": 32, |
| 63 | "num_return_sequences": 1, |
| 64 | "out_features": [ |
| 65 | "stage1", |
| 66 | "stage2", |
| 67 | "stage3" |
| 68 | ], |
| 69 | "output_attentions": false, |
| 70 | "output_hidden_states": false, |
| 71 | "output_scores": false, |
| 72 | "output_stride": 32, |
| 73 | "pad_token_id": null, |
| 74 | "prefix": null, |
| 75 | "problem_type": null, |
| 76 | "pruned_heads": {}, |
| 77 | "remove_invalid_values": false, |
| 78 | "repetition_penalty": 1.0, |
| 79 | "return_dict": true, |
| 80 | "return_dict_in_generate": false, |
| 81 | "sep_token_id": null, |
| 82 | "stage_names": [ |
| 83 | "stem", |
| 84 | "stage1", |
| 85 | "stage2", |
| 86 | "stage3" |
| 87 | ], |
| 88 | "suppress_tokens": null, |
| 89 | "task_specific_params": null, |
| 90 | "temperature": 1.0, |
| 91 | "tf_legacy_loss": false, |
| 92 | "tie_encoder_decoder": false, |
| 93 | "tie_word_embeddings": true, |
| 94 | "tokenizer_class": null, |
| 95 | "top_k": 50, |
| 96 | "top_p": 1.0, |
| 97 | "torch_dtype": null, |
| 98 | "torchscript": false, |
| 99 | "transformers_version": "4.26.0.dev0", |
| 100 | "typical_p": 1.0, |
| 101 | "use_bfloat16": false, |
| 102 | "width_factor": 1 |
| 103 | }, |
| 104 | "backbone_featmap_shape": [ |
| 105 | 1, |
| 106 | 1024, |
| 107 | 24, |
| 108 | 24 |
| 109 | ], |
| 110 | "backbone_out_indices": [ |
| 111 | 2, |
| 112 | 5, |
| 113 | 8, |
| 114 | 11 |
| 115 | ], |
| 116 | "fusion_hidden_size": 256, |
| 117 | "head_in_index": -1, |
| 118 | "hidden_act": "gelu", |
| 119 | "hidden_dropout_prob": 0.0, |
| 120 | "hidden_size": 768, |
| 121 | "id2label": { |
| 122 | "0": "LABEL_0", |
| 123 | "1": "LABEL_1", |
| 124 | "2": "LABEL_2", |
| 125 | "3": "LABEL_3", |
| 126 | "4": "LABEL_4", |
| 127 | "5": "LABEL_5", |
| 128 | "6": "LABEL_6", |
| 129 | "7": "LABEL_7", |
| 130 | "8": "LABEL_8", |
| 131 | "9": "LABEL_9", |
| 132 | "10": "LABEL_10", |
| 133 | "11": "LABEL_11", |
| 134 | "12": "LABEL_12", |
| 135 | "13": "LABEL_13", |
| 136 | "14": "LABEL_14", |
| 137 | "15": "LABEL_15", |
| 138 | "16": "LABEL_16", |
| 139 | "17": "LABEL_17", |
| 140 | "18": "LABEL_18", |
| 141 | "19": "LABEL_19", |
| 142 | "20": "LABEL_20", |
| 143 | "21": "LABEL_21", |
| 144 | "22": "LABEL_22", |
| 145 | "23": "LABEL_23", |
| 146 | "24": "LABEL_24", |
| 147 | "25": "LABEL_25", |
| 148 | "26": "LABEL_26", |
| 149 | "27": "LABEL_27", |
| 150 | "28": "LABEL_28", |
| 151 | "29": "LABEL_29", |
| 152 | "30": "LABEL_30", |
| 153 | "31": "LABEL_31", |
| 154 | "32": "LABEL_32", |
| 155 | "33": "LABEL_33", |
| 156 | "34": "LABEL_34", |
| 157 | "35": "LABEL_35", |
| 158 | "36": "LABEL_36", |
| 159 | "37": "LABEL_37", |
| 160 | "38": "LABEL_38", |
| 161 | "39": "LABEL_39", |
| 162 | "40": "LABEL_40", |
| 163 | "41": "LABEL_41", |
| 164 | "42": "LABEL_42", |
| 165 | "43": "LABEL_43", |
| 166 | "44": "LABEL_44", |
| 167 | "45": "LABEL_45", |
| 168 | "46": "LABEL_46", |
| 169 | "47": "LABEL_47", |
| 170 | "48": "LABEL_48", |
| 171 | "49": "LABEL_49", |
| 172 | "50": "LABEL_50", |
| 173 | "51": "LABEL_51", |
| 174 | "52": "LABEL_52", |
| 175 | "53": "LABEL_53", |
| 176 | "54": "LABEL_54", |
| 177 | "55": "LABEL_55", |
| 178 | "56": "LABEL_56", |
| 179 | "57": "LABEL_57", |
| 180 | "58": "LABEL_58", |
| 181 | "59": "LABEL_59", |
| 182 | "60": "LABEL_60", |
| 183 | "61": "LABEL_61", |
| 184 | "62": "LABEL_62", |
| 185 | "63": "LABEL_63", |
| 186 | "64": "LABEL_64", |
| 187 | "65": "LABEL_65", |
| 188 | "66": "LABEL_66", |
| 189 | "67": "LABEL_67", |
| 190 | "68": "LABEL_68", |
| 191 | "69": "LABEL_69", |
| 192 | "70": "LABEL_70", |
| 193 | "71": "LABEL_71", |
| 194 | "72": "LABEL_72", |
| 195 | "73": "LABEL_73", |
| 196 | "74": "LABEL_74", |
| 197 | "75": "LABEL_75", |
| 198 | "76": "LABEL_76", |
| 199 | "77": "LABEL_77", |
| 200 | "78": "LABEL_78", |
| 201 | "79": "LABEL_79", |
| 202 | "80": "LABEL_80", |
| 203 | "81": "LABEL_81", |
| 204 | "82": "LABEL_82", |
| 205 | "83": "LABEL_83", |
| 206 | "84": "LABEL_84", |
| 207 | "85": "LABEL_85", |
| 208 | "86": "LABEL_86", |
| 209 | "87": "LABEL_87", |
| 210 | "88": "LABEL_88", |
| 211 | "89": "LABEL_89", |
| 212 | "90": "LABEL_90", |
| 213 | "91": "LABEL_91", |
| 214 | "92": "LABEL_92", |
| 215 | "93": "LABEL_93", |
| 216 | "94": "LABEL_94", |
| 217 | "95": "LABEL_95", |
| 218 | "96": "LABEL_96", |
| 219 | "97": "LABEL_97", |
| 220 | "98": "LABEL_98", |
| 221 | "99": "LABEL_99", |
| 222 | "100": "LABEL_100", |
| 223 | "101": "LABEL_101", |
| 224 | "102": "LABEL_102", |
| 225 | "103": "LABEL_103", |
| 226 | "104": "LABEL_104", |
| 227 | "105": "LABEL_105", |
| 228 | "106": "LABEL_106", |
| 229 | "107": "LABEL_107", |
| 230 | "108": "LABEL_108", |
| 231 | "109": "LABEL_109", |
| 232 | "110": "LABEL_110", |
| 233 | "111": "LABEL_111", |
| 234 | "112": "LABEL_112", |
| 235 | "113": "LABEL_113", |
| 236 | "114": "LABEL_114", |
| 237 | "115": "LABEL_115", |
| 238 | "116": "LABEL_116", |
| 239 | "117": "LABEL_117", |
| 240 | "118": "LABEL_118", |
| 241 | "119": "LABEL_119", |
| 242 | "120": "LABEL_120", |
| 243 | "121": "LABEL_121", |
| 244 | "122": "LABEL_122", |
| 245 | "123": "LABEL_123", |
| 246 | "124": "LABEL_124", |
| 247 | "125": "LABEL_125", |
| 248 | "126": "LABEL_126", |
| 249 | "127": "LABEL_127", |
| 250 | "128": "LABEL_128", |
| 251 | "129": "LABEL_129", |
| 252 | "130": "LABEL_130", |
| 253 | "131": "LABEL_131", |
| 254 | "132": "LABEL_132", |
| 255 | "133": "LABEL_133", |
| 256 | "134": "LABEL_134", |
| 257 | "135": "LABEL_135", |
| 258 | "136": "LABEL_136", |
| 259 | "137": "LABEL_137", |
| 260 | "138": "LABEL_138", |
| 261 | "139": "LABEL_139", |
| 262 | "140": "LABEL_140", |
| 263 | "141": "LABEL_141", |
| 264 | "142": "LABEL_142", |
| 265 | "143": "LABEL_143", |
| 266 | "144": "LABEL_144", |
| 267 | "145": "LABEL_145", |
| 268 | "146": "LABEL_146", |
| 269 | "147": "LABEL_147", |
| 270 | "148": "LABEL_148", |
| 271 | "149": "LABEL_149" |
| 272 | }, |
| 273 | "image_size": 384, |
| 274 | "initializer_range": 0.02, |
| 275 | "intermediate_size": 3072, |
| 276 | "is_hybrid": true, |
| 277 | "label2id": { |
| 278 | "LABEL_0": 0, |
| 279 | "LABEL_1": 1, |
| 280 | "LABEL_10": 10, |
| 281 | "LABEL_100": 100, |
| 282 | "LABEL_101": 101, |
| 283 | "LABEL_102": 102, |
| 284 | "LABEL_103": 103, |
| 285 | "LABEL_104": 104, |
| 286 | "LABEL_105": 105, |
| 287 | "LABEL_106": 106, |
| 288 | "LABEL_107": 107, |
| 289 | "LABEL_108": 108, |
| 290 | "LABEL_109": 109, |
| 291 | "LABEL_11": 11, |
| 292 | "LABEL_110": 110, |
| 293 | "LABEL_111": 111, |
| 294 | "LABEL_112": 112, |
| 295 | "LABEL_113": 113, |
| 296 | "LABEL_114": 114, |
| 297 | "LABEL_115": 115, |
| 298 | "LABEL_116": 116, |
| 299 | "LABEL_117": 117, |
| 300 | "LABEL_118": 118, |
| 301 | "LABEL_119": 119, |
| 302 | "LABEL_12": 12, |
| 303 | "LABEL_120": 120, |
| 304 | "LABEL_121": 121, |
| 305 | "LABEL_122": 122, |
| 306 | "LABEL_123": 123, |
| 307 | "LABEL_124": 124, |
| 308 | "LABEL_125": 125, |
| 309 | "LABEL_126": 126, |
| 310 | "LABEL_127": 127, |
| 311 | "LABEL_128": 128, |
| 312 | "LABEL_129": 129, |
| 313 | "LABEL_13": 13, |
| 314 | "LABEL_130": 130, |
| 315 | "LABEL_131": 131, |
| 316 | "LABEL_132": 132, |
| 317 | "LABEL_133": 133, |
| 318 | "LABEL_134": 134, |
| 319 | "LABEL_135": 135, |
| 320 | "LABEL_136": 136, |
| 321 | "LABEL_137": 137, |
| 322 | "LABEL_138": 138, |
| 323 | "LABEL_139": 139, |
| 324 | "LABEL_14": 14, |
| 325 | "LABEL_140": 140, |
| 326 | "LABEL_141": 141, |
| 327 | "LABEL_142": 142, |
| 328 | "LABEL_143": 143, |
| 329 | "LABEL_144": 144, |
| 330 | "LABEL_145": 145, |
| 331 | "LABEL_146": 146, |
| 332 | "LABEL_147": 147, |
| 333 | "LABEL_148": 148, |
| 334 | "LABEL_149": 149, |
| 335 | "LABEL_15": 15, |
| 336 | "LABEL_16": 16, |
| 337 | "LABEL_17": 17, |
| 338 | "LABEL_18": 18, |
| 339 | "LABEL_19": 19, |
| 340 | "LABEL_2": 2, |
| 341 | "LABEL_20": 20, |
| 342 | "LABEL_21": 21, |
| 343 | "LABEL_22": 22, |
| 344 | "LABEL_23": 23, |
| 345 | "LABEL_24": 24, |
| 346 | "LABEL_25": 25, |
| 347 | "LABEL_26": 26, |
| 348 | "LABEL_27": 27, |
| 349 | "LABEL_28": 28, |
| 350 | "LABEL_29": 29, |
| 351 | "LABEL_3": 3, |
| 352 | "LABEL_30": 30, |
| 353 | "LABEL_31": 31, |
| 354 | "LABEL_32": 32, |
| 355 | "LABEL_33": 33, |
| 356 | "LABEL_34": 34, |
| 357 | "LABEL_35": 35, |
| 358 | "LABEL_36": 36, |
| 359 | "LABEL_37": 37, |
| 360 | "LABEL_38": 38, |
| 361 | "LABEL_39": 39, |
| 362 | "LABEL_4": 4, |
| 363 | "LABEL_40": 40, |
| 364 | "LABEL_41": 41, |
| 365 | "LABEL_42": 42, |
| 366 | "LABEL_43": 43, |
| 367 | "LABEL_44": 44, |
| 368 | "LABEL_45": 45, |
| 369 | "LABEL_46": 46, |
| 370 | "LABEL_47": 47, |
| 371 | "LABEL_48": 48, |
| 372 | "LABEL_49": 49, |
| 373 | "LABEL_5": 5, |
| 374 | "LABEL_50": 50, |
| 375 | "LABEL_51": 51, |
| 376 | "LABEL_52": 52, |
| 377 | "LABEL_53": 53, |
| 378 | "LABEL_54": 54, |
| 379 | "LABEL_55": 55, |
| 380 | "LABEL_56": 56, |
| 381 | "LABEL_57": 57, |
| 382 | "LABEL_58": 58, |
| 383 | "LABEL_59": 59, |
| 384 | "LABEL_6": 6, |
| 385 | "LABEL_60": 60, |
| 386 | "LABEL_61": 61, |
| 387 | "LABEL_62": 62, |
| 388 | "LABEL_63": 63, |
| 389 | "LABEL_64": 64, |
| 390 | "LABEL_65": 65, |
| 391 | "LABEL_66": 66, |
| 392 | "LABEL_67": 67, |
| 393 | "LABEL_68": 68, |
| 394 | "LABEL_69": 69, |
| 395 | "LABEL_7": 7, |
| 396 | "LABEL_70": 70, |
| 397 | "LABEL_71": 71, |
| 398 | "LABEL_72": 72, |
| 399 | "LABEL_73": 73, |
| 400 | "LABEL_74": 74, |
| 401 | "LABEL_75": 75, |
| 402 | "LABEL_76": 76, |
| 403 | "LABEL_77": 77, |
| 404 | "LABEL_78": 78, |
| 405 | "LABEL_79": 79, |
| 406 | "LABEL_8": 8, |
| 407 | "LABEL_80": 80, |
| 408 | "LABEL_81": 81, |
| 409 | "LABEL_82": 82, |
| 410 | "LABEL_83": 83, |
| 411 | "LABEL_84": 84, |
| 412 | "LABEL_85": 85, |
| 413 | "LABEL_86": 86, |
| 414 | "LABEL_87": 87, |
| 415 | "LABEL_88": 88, |
| 416 | "LABEL_89": 89, |
| 417 | "LABEL_9": 9, |
| 418 | "LABEL_90": 90, |
| 419 | "LABEL_91": 91, |
| 420 | "LABEL_92": 92, |
| 421 | "LABEL_93": 93, |
| 422 | "LABEL_94": 94, |
| 423 | "LABEL_95": 95, |
| 424 | "LABEL_96": 96, |
| 425 | "LABEL_97": 97, |
| 426 | "LABEL_98": 98, |
| 427 | "LABEL_99": 99 |
| 428 | }, |
| 429 | "layer_norm_eps": 1e-12, |
| 430 | "model_type": "dpt", |
| 431 | "neck_hidden_sizes": [ |
| 432 | 256, |
| 433 | 512, |
| 434 | 768, |
| 435 | 768 |
| 436 | ], |
| 437 | "neck_ignore_stages": [ |
| 438 | 0, |
| 439 | 1 |
| 440 | ], |
| 441 | "num_attention_heads": 12, |
| 442 | "num_channels": 3, |
| 443 | "num_hidden_layers": 12, |
| 444 | "patch_size": 16, |
| 445 | "qkv_bias": true, |
| 446 | "readout_type": "project", |
| 447 | "reassemble_factors": [ |
| 448 | 1, |
| 449 | 1, |
| 450 | 1, |
| 451 | 0.5 |
| 452 | ], |
| 453 | "semantic_classifier_dropout": 0.1, |
| 454 | "semantic_loss_ignore_index": 255, |
| 455 | "torch_dtype": "float32", |
| 456 | "transformers_version": null, |
| 457 | "use_auxiliary_head": true, |
| 458 | "use_batch_norm_in_fusion_residual": false |
| 459 | } |
| 460 | |