Training in progress, epoch 1, checkpoint
Browse files- checkpoint-740/config.json +351 -0
- checkpoint-740/model.safetensors +3 -0
- checkpoint-740/optimizer.pt +3 -0
- checkpoint-740/preprocessor_config.json +36 -0
- checkpoint-740/rng_state.pth +3 -0
- checkpoint-740/scaler.pt +3 -0
- checkpoint-740/scheduler.pt +3 -0
- checkpoint-740/trainer_state.json +219 -0
- checkpoint-740/training_args.bin +3 -0
checkpoint-740/config.json
ADDED
|
@@ -0,0 +1,351 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation_dropout": 0.0,
|
| 3 |
+
"activation_function": "silu",
|
| 4 |
+
"anchor_image_size": null,
|
| 5 |
+
"architectures": [
|
| 6 |
+
"DFineForObjectDetection"
|
| 7 |
+
],
|
| 8 |
+
"attention_dropout": 0.0,
|
| 9 |
+
"auxiliary_loss": true,
|
| 10 |
+
"backbone": null,
|
| 11 |
+
"backbone_config": {
|
| 12 |
+
"depths": [
|
| 13 |
+
3,
|
| 14 |
+
4,
|
| 15 |
+
6,
|
| 16 |
+
3
|
| 17 |
+
],
|
| 18 |
+
"downsample_in_bottleneck": false,
|
| 19 |
+
"downsample_in_first_stage": false,
|
| 20 |
+
"dtype": "float32",
|
| 21 |
+
"embedding_size": 32,
|
| 22 |
+
"hidden_act": "relu",
|
| 23 |
+
"hidden_sizes": [
|
| 24 |
+
256,
|
| 25 |
+
512,
|
| 26 |
+
1024,
|
| 27 |
+
2048
|
| 28 |
+
],
|
| 29 |
+
"initializer_range": 0.02,
|
| 30 |
+
"layer_type": "basic",
|
| 31 |
+
"model_type": "hgnet_v2",
|
| 32 |
+
"num_channels": 3,
|
| 33 |
+
"out_features": [
|
| 34 |
+
"stage2",
|
| 35 |
+
"stage3",
|
| 36 |
+
"stage4"
|
| 37 |
+
],
|
| 38 |
+
"out_indices": [
|
| 39 |
+
2,
|
| 40 |
+
3,
|
| 41 |
+
4
|
| 42 |
+
],
|
| 43 |
+
"stage_downsample": [
|
| 44 |
+
false,
|
| 45 |
+
true,
|
| 46 |
+
true,
|
| 47 |
+
true
|
| 48 |
+
],
|
| 49 |
+
"stage_in_channels": [
|
| 50 |
+
48,
|
| 51 |
+
128,
|
| 52 |
+
512,
|
| 53 |
+
1024
|
| 54 |
+
],
|
| 55 |
+
"stage_kernel_size": [
|
| 56 |
+
3,
|
| 57 |
+
3,
|
| 58 |
+
5,
|
| 59 |
+
5
|
| 60 |
+
],
|
| 61 |
+
"stage_light_block": [
|
| 62 |
+
false,
|
| 63 |
+
false,
|
| 64 |
+
true,
|
| 65 |
+
true
|
| 66 |
+
],
|
| 67 |
+
"stage_mid_channels": [
|
| 68 |
+
48,
|
| 69 |
+
96,
|
| 70 |
+
192,
|
| 71 |
+
384
|
| 72 |
+
],
|
| 73 |
+
"stage_names": [
|
| 74 |
+
"stem",
|
| 75 |
+
"stage1",
|
| 76 |
+
"stage2",
|
| 77 |
+
"stage3",
|
| 78 |
+
"stage4"
|
| 79 |
+
],
|
| 80 |
+
"stage_num_blocks": [
|
| 81 |
+
1,
|
| 82 |
+
1,
|
| 83 |
+
3,
|
| 84 |
+
1
|
| 85 |
+
],
|
| 86 |
+
"stage_numb_of_layers": [
|
| 87 |
+
6,
|
| 88 |
+
6,
|
| 89 |
+
6,
|
| 90 |
+
6
|
| 91 |
+
],
|
| 92 |
+
"stage_out_channels": [
|
| 93 |
+
128,
|
| 94 |
+
512,
|
| 95 |
+
1024,
|
| 96 |
+
2048
|
| 97 |
+
],
|
| 98 |
+
"stem_channels": [
|
| 99 |
+
3,
|
| 100 |
+
32,
|
| 101 |
+
48
|
| 102 |
+
],
|
| 103 |
+
"use_learnable_affine_block": false
|
| 104 |
+
},
|
| 105 |
+
"backbone_kwargs": null,
|
| 106 |
+
"batch_norm_eps": 1e-05,
|
| 107 |
+
"box_noise_scale": 1.0,
|
| 108 |
+
"d_model": 256,
|
| 109 |
+
"decoder_activation_function": "relu",
|
| 110 |
+
"decoder_attention_heads": 8,
|
| 111 |
+
"decoder_ffn_dim": 1024,
|
| 112 |
+
"decoder_in_channels": [
|
| 113 |
+
256,
|
| 114 |
+
256,
|
| 115 |
+
256
|
| 116 |
+
],
|
| 117 |
+
"decoder_layers": 6,
|
| 118 |
+
"decoder_method": "default",
|
| 119 |
+
"decoder_n_points": [
|
| 120 |
+
3,
|
| 121 |
+
6,
|
| 122 |
+
3
|
| 123 |
+
],
|
| 124 |
+
"decoder_offset_scale": 0.5,
|
| 125 |
+
"depth_mult": 1.0,
|
| 126 |
+
"dropout": 0.0,
|
| 127 |
+
"dtype": "float32",
|
| 128 |
+
"encode_proj_layers": [
|
| 129 |
+
2
|
| 130 |
+
],
|
| 131 |
+
"encoder_activation_function": "gelu",
|
| 132 |
+
"encoder_attention_heads": 8,
|
| 133 |
+
"encoder_ffn_dim": 1024,
|
| 134 |
+
"encoder_hidden_dim": 256,
|
| 135 |
+
"encoder_in_channels": [
|
| 136 |
+
512,
|
| 137 |
+
1024,
|
| 138 |
+
2048
|
| 139 |
+
],
|
| 140 |
+
"encoder_layers": 1,
|
| 141 |
+
"eos_coefficient": 0.0001,
|
| 142 |
+
"eval_idx": -1,
|
| 143 |
+
"eval_size": null,
|
| 144 |
+
"feat_strides": [
|
| 145 |
+
8,
|
| 146 |
+
16,
|
| 147 |
+
32
|
| 148 |
+
],
|
| 149 |
+
"focal_loss_alpha": 0.75,
|
| 150 |
+
"focal_loss_gamma": 2.0,
|
| 151 |
+
"freeze_backbone_batch_norms": true,
|
| 152 |
+
"hidden_expansion": 1.0,
|
| 153 |
+
"id2label": {
|
| 154 |
+
"0": "person",
|
| 155 |
+
"1": "bicycle",
|
| 156 |
+
"2": "car",
|
| 157 |
+
"3": "motorcycle",
|
| 158 |
+
"4": "airplane",
|
| 159 |
+
"5": "bus",
|
| 160 |
+
"6": "train",
|
| 161 |
+
"7": "truck",
|
| 162 |
+
"8": "boat",
|
| 163 |
+
"9": "traffic light",
|
| 164 |
+
"10": "fire hydrant",
|
| 165 |
+
"11": "stop sign",
|
| 166 |
+
"12": "parking meter",
|
| 167 |
+
"13": "bench",
|
| 168 |
+
"14": "bird",
|
| 169 |
+
"15": "cat",
|
| 170 |
+
"16": "dog",
|
| 171 |
+
"17": "horse",
|
| 172 |
+
"18": "sheep",
|
| 173 |
+
"19": "cow",
|
| 174 |
+
"20": "elephant",
|
| 175 |
+
"21": "bear",
|
| 176 |
+
"22": "zebra",
|
| 177 |
+
"23": "giraffe",
|
| 178 |
+
"24": "backpack",
|
| 179 |
+
"25": "umbrella",
|
| 180 |
+
"26": "handbag",
|
| 181 |
+
"27": "tie",
|
| 182 |
+
"28": "suitcase",
|
| 183 |
+
"29": "frisbee",
|
| 184 |
+
"30": "skis",
|
| 185 |
+
"31": "snowboard",
|
| 186 |
+
"32": "sports ball",
|
| 187 |
+
"33": "kite",
|
| 188 |
+
"34": "baseball bat",
|
| 189 |
+
"35": "baseball glove",
|
| 190 |
+
"36": "skateboard",
|
| 191 |
+
"37": "surfboard",
|
| 192 |
+
"38": "tennis racket",
|
| 193 |
+
"39": "bottle",
|
| 194 |
+
"40": "wine glass",
|
| 195 |
+
"41": "cup",
|
| 196 |
+
"42": "fork",
|
| 197 |
+
"43": "knife",
|
| 198 |
+
"44": "spoon",
|
| 199 |
+
"45": "bowl",
|
| 200 |
+
"46": "banana",
|
| 201 |
+
"47": "apple",
|
| 202 |
+
"48": "sandwich",
|
| 203 |
+
"49": "orange",
|
| 204 |
+
"50": "broccoli",
|
| 205 |
+
"51": "carrot",
|
| 206 |
+
"52": "hot dog",
|
| 207 |
+
"53": "pizza",
|
| 208 |
+
"54": "donut",
|
| 209 |
+
"55": "cake",
|
| 210 |
+
"56": "chair",
|
| 211 |
+
"57": "couch",
|
| 212 |
+
"58": "potted plant",
|
| 213 |
+
"59": "bed",
|
| 214 |
+
"60": "dining table",
|
| 215 |
+
"61": "toilet",
|
| 216 |
+
"62": "tv",
|
| 217 |
+
"63": "laptop",
|
| 218 |
+
"64": "mouse",
|
| 219 |
+
"65": "remote",
|
| 220 |
+
"66": "keyboard",
|
| 221 |
+
"67": "cell phone",
|
| 222 |
+
"68": "microwave",
|
| 223 |
+
"69": "oven",
|
| 224 |
+
"70": "toaster",
|
| 225 |
+
"71": "sink",
|
| 226 |
+
"72": "refrigerator",
|
| 227 |
+
"73": "book",
|
| 228 |
+
"74": "clock",
|
| 229 |
+
"75": "vase",
|
| 230 |
+
"76": "scissors",
|
| 231 |
+
"77": "teddy bear",
|
| 232 |
+
"78": "hair drier",
|
| 233 |
+
"79": "toothbrush"
|
| 234 |
+
},
|
| 235 |
+
"initializer_bias_prior_prob": null,
|
| 236 |
+
"initializer_range": 0.01,
|
| 237 |
+
"is_encoder_decoder": true,
|
| 238 |
+
"label2id": {
|
| 239 |
+
"airplane": 4,
|
| 240 |
+
"apple": 47,
|
| 241 |
+
"backpack": 24,
|
| 242 |
+
"banana": 46,
|
| 243 |
+
"baseball bat": 34,
|
| 244 |
+
"baseball glove": 35,
|
| 245 |
+
"bear": 21,
|
| 246 |
+
"bed": 59,
|
| 247 |
+
"bench": 13,
|
| 248 |
+
"bicycle": 1,
|
| 249 |
+
"bird": 14,
|
| 250 |
+
"boat": 8,
|
| 251 |
+
"book": 73,
|
| 252 |
+
"bottle": 39,
|
| 253 |
+
"bowl": 45,
|
| 254 |
+
"broccoli": 50,
|
| 255 |
+
"bus": 5,
|
| 256 |
+
"cake": 55,
|
| 257 |
+
"car": 2,
|
| 258 |
+
"carrot": 51,
|
| 259 |
+
"cat": 15,
|
| 260 |
+
"cell phone": 67,
|
| 261 |
+
"chair": 56,
|
| 262 |
+
"clock": 74,
|
| 263 |
+
"couch": 57,
|
| 264 |
+
"cow": 19,
|
| 265 |
+
"cup": 41,
|
| 266 |
+
"dining table": 60,
|
| 267 |
+
"dog": 16,
|
| 268 |
+
"donut": 54,
|
| 269 |
+
"elephant": 20,
|
| 270 |
+
"fire hydrant": 10,
|
| 271 |
+
"fork": 42,
|
| 272 |
+
"frisbee": 29,
|
| 273 |
+
"giraffe": 23,
|
| 274 |
+
"hair drier": 78,
|
| 275 |
+
"handbag": 26,
|
| 276 |
+
"horse": 17,
|
| 277 |
+
"hot dog": 52,
|
| 278 |
+
"keyboard": 66,
|
| 279 |
+
"kite": 33,
|
| 280 |
+
"knife": 43,
|
| 281 |
+
"laptop": 63,
|
| 282 |
+
"microwave": 68,
|
| 283 |
+
"motorcycle": 3,
|
| 284 |
+
"mouse": 64,
|
| 285 |
+
"orange": 49,
|
| 286 |
+
"oven": 69,
|
| 287 |
+
"parking meter": 12,
|
| 288 |
+
"person": 0,
|
| 289 |
+
"pizza": 53,
|
| 290 |
+
"potted plant": 58,
|
| 291 |
+
"refrigerator": 72,
|
| 292 |
+
"remote": 65,
|
| 293 |
+
"sandwich": 48,
|
| 294 |
+
"scissors": 76,
|
| 295 |
+
"sheep": 18,
|
| 296 |
+
"sink": 71,
|
| 297 |
+
"skateboard": 36,
|
| 298 |
+
"skis": 30,
|
| 299 |
+
"snowboard": 31,
|
| 300 |
+
"spoon": 44,
|
| 301 |
+
"sports ball": 32,
|
| 302 |
+
"stop sign": 11,
|
| 303 |
+
"suitcase": 28,
|
| 304 |
+
"surfboard": 37,
|
| 305 |
+
"teddy bear": 77,
|
| 306 |
+
"tennis racket": 38,
|
| 307 |
+
"tie": 27,
|
| 308 |
+
"toaster": 70,
|
| 309 |
+
"toilet": 61,
|
| 310 |
+
"toothbrush": 79,
|
| 311 |
+
"traffic light": 9,
|
| 312 |
+
"train": 6,
|
| 313 |
+
"truck": 7,
|
| 314 |
+
"tv": 62,
|
| 315 |
+
"umbrella": 25,
|
| 316 |
+
"vase": 75,
|
| 317 |
+
"wine glass": 40,
|
| 318 |
+
"zebra": 22
|
| 319 |
+
},
|
| 320 |
+
"label_noise_ratio": 0.5,
|
| 321 |
+
"layer_norm_eps": 1e-05,
|
| 322 |
+
"layer_scale": 1,
|
| 323 |
+
"learn_initial_query": false,
|
| 324 |
+
"lqe_hidden_dim": 64,
|
| 325 |
+
"lqe_layers": 2,
|
| 326 |
+
"matcher_alpha": 0.25,
|
| 327 |
+
"matcher_bbox_cost": 5.0,
|
| 328 |
+
"matcher_class_cost": 2.0,
|
| 329 |
+
"matcher_gamma": 2.0,
|
| 330 |
+
"matcher_giou_cost": 2.0,
|
| 331 |
+
"max_num_bins": 32,
|
| 332 |
+
"model_type": "d_fine",
|
| 333 |
+
"normalize_before": false,
|
| 334 |
+
"num_denoising": 100,
|
| 335 |
+
"num_feature_levels": 3,
|
| 336 |
+
"num_queries": 300,
|
| 337 |
+
"positional_encoding_temperature": 10000,
|
| 338 |
+
"reg_scale": 4.0,
|
| 339 |
+
"top_prob_values": 4,
|
| 340 |
+
"transformers_version": "4.57.3",
|
| 341 |
+
"up": 0.5,
|
| 342 |
+
"use_focal_loss": true,
|
| 343 |
+
"use_pretrained_backbone": false,
|
| 344 |
+
"use_timm_backbone": false,
|
| 345 |
+
"weight_loss_bbox": 5.0,
|
| 346 |
+
"weight_loss_ddf": 1.5,
|
| 347 |
+
"weight_loss_fgl": 0.15,
|
| 348 |
+
"weight_loss_giou": 2.0,
|
| 349 |
+
"weight_loss_vfl": 1.0,
|
| 350 |
+
"with_box_refine": true
|
| 351 |
+
}
|
checkpoint-740/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:33444fe7aef107e1465667197fb4ccedaef3b482e3dcac1b3f4b609f6c299074
|
| 3 |
+
size 125618504
|
checkpoint-740/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b15bc489560dcb85f7ad702cfa3b251c8bee3ad2423e99e733b756be2e9fea25
|
| 3 |
+
size 250468721
|
checkpoint-740/preprocessor_config.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"crop_size": null,
|
| 3 |
+
"data_format": "channels_first",
|
| 4 |
+
"default_to_square": false,
|
| 5 |
+
"device": null,
|
| 6 |
+
"disable_grouping": null,
|
| 7 |
+
"do_center_crop": null,
|
| 8 |
+
"do_convert_annotations": true,
|
| 9 |
+
"do_convert_rgb": null,
|
| 10 |
+
"do_normalize": false,
|
| 11 |
+
"do_pad": false,
|
| 12 |
+
"do_rescale": true,
|
| 13 |
+
"do_resize": true,
|
| 14 |
+
"format": "coco_detection",
|
| 15 |
+
"image_mean": [
|
| 16 |
+
0.485,
|
| 17 |
+
0.456,
|
| 18 |
+
0.406
|
| 19 |
+
],
|
| 20 |
+
"image_processor_type": "RTDetrImageProcessorFast",
|
| 21 |
+
"image_std": [
|
| 22 |
+
0.229,
|
| 23 |
+
0.224,
|
| 24 |
+
0.225
|
| 25 |
+
],
|
| 26 |
+
"input_data_format": null,
|
| 27 |
+
"pad_size": null,
|
| 28 |
+
"resample": 2,
|
| 29 |
+
"rescale_factor": 0.00392156862745098,
|
| 30 |
+
"return_segmentation_masks": null,
|
| 31 |
+
"return_tensors": null,
|
| 32 |
+
"size": {
|
| 33 |
+
"height": 832,
|
| 34 |
+
"width": 832
|
| 35 |
+
}
|
| 36 |
+
}
|
checkpoint-740/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ffde4745484f100270f6018dc66aaba0309fe2d7e8609ae0ba9062e8a552d4d
|
| 3 |
+
size 14645
|
checkpoint-740/scaler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9cc43bdc3b64e2c5dc243a0f45294a998ed29f94ec64d3bf06b306513f5023cd
|
| 3 |
+
size 1383
|
checkpoint-740/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cbc97741995f07213194ca8e6fc216c359c7f2b7ae6a4c57c97cd3360095f899
|
| 3 |
+
size 1465
|
checkpoint-740/trainer_state.json
ADDED
|
@@ -0,0 +1,219 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 740,
|
| 3 |
+
"best_metric": 0.0299,
|
| 4 |
+
"best_model_checkpoint": "output/checkpoint-740",
|
| 5 |
+
"epoch": 1.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 740,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.6756756756756757,
|
| 14 |
+
"grad_norm": 178.42633056640625,
|
| 15 |
+
"learning_rate": 2.7386363636363637e-05,
|
| 16 |
+
"loss": 76.1385,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 1.0,
|
| 21 |
+
"eval_loss": 3.011831521987915,
|
| 22 |
+
"eval_map": 0.0299,
|
| 23 |
+
"eval_map_50": 0.0372,
|
| 24 |
+
"eval_map_75": 0.0329,
|
| 25 |
+
"eval_map_airplane": 0.0221,
|
| 26 |
+
"eval_map_apple": 0.0,
|
| 27 |
+
"eval_map_backpack": 0.0021,
|
| 28 |
+
"eval_map_banana": 0.0063,
|
| 29 |
+
"eval_map_baseball bat": 0.0,
|
| 30 |
+
"eval_map_baseball glove": 0.1984,
|
| 31 |
+
"eval_map_bear": 0.0,
|
| 32 |
+
"eval_map_bed": 0.0,
|
| 33 |
+
"eval_map_bench": 0.0009,
|
| 34 |
+
"eval_map_bicycle": 0.0,
|
| 35 |
+
"eval_map_bird": 0.0016,
|
| 36 |
+
"eval_map_boat": 0.0123,
|
| 37 |
+
"eval_map_book": 0.0002,
|
| 38 |
+
"eval_map_bottle": 0.046,
|
| 39 |
+
"eval_map_bowl": 0.0588,
|
| 40 |
+
"eval_map_broccoli": 0.0007,
|
| 41 |
+
"eval_map_bus": 0.3589,
|
| 42 |
+
"eval_map_cake": 0.0142,
|
| 43 |
+
"eval_map_car": 0.1625,
|
| 44 |
+
"eval_map_carrot": 0.0,
|
| 45 |
+
"eval_map_cat": 0.0047,
|
| 46 |
+
"eval_map_cell phone": 0.08,
|
| 47 |
+
"eval_map_chair": 0.0741,
|
| 48 |
+
"eval_map_clock": 0.001,
|
| 49 |
+
"eval_map_couch": 0.0125,
|
| 50 |
+
"eval_map_cow": 0.1326,
|
| 51 |
+
"eval_map_cup": 0.0592,
|
| 52 |
+
"eval_map_dining table": 0.0596,
|
| 53 |
+
"eval_map_dog": 0.0,
|
| 54 |
+
"eval_map_donut": -1.0,
|
| 55 |
+
"eval_map_elephant": 0.0034,
|
| 56 |
+
"eval_map_fire hydrant": -1.0,
|
| 57 |
+
"eval_map_fork": 0.0027,
|
| 58 |
+
"eval_map_frisbee": 0.0008,
|
| 59 |
+
"eval_map_giraffe": 0.0976,
|
| 60 |
+
"eval_map_handbag": 0.0022,
|
| 61 |
+
"eval_map_horse": 0.0022,
|
| 62 |
+
"eval_map_hot dog": 0.0,
|
| 63 |
+
"eval_map_keyboard": 0.0,
|
| 64 |
+
"eval_map_kite": 0.008,
|
| 65 |
+
"eval_map_knife": 0.001,
|
| 66 |
+
"eval_map_laptop": 0.0009,
|
| 67 |
+
"eval_map_large": 0.0435,
|
| 68 |
+
"eval_map_medium": 0.0369,
|
| 69 |
+
"eval_map_microwave": 0.0,
|
| 70 |
+
"eval_map_motorcycle": 0.0006,
|
| 71 |
+
"eval_map_mouse": 0.0,
|
| 72 |
+
"eval_map_orange": 0.0,
|
| 73 |
+
"eval_map_oven": 0.0,
|
| 74 |
+
"eval_map_parking meter": 0.0,
|
| 75 |
+
"eval_map_person": 0.1633,
|
| 76 |
+
"eval_map_pizza": 0.0451,
|
| 77 |
+
"eval_map_potted plant": 0.0026,
|
| 78 |
+
"eval_map_refrigerator": 0.0013,
|
| 79 |
+
"eval_map_remote": 0.0053,
|
| 80 |
+
"eval_map_sandwich": 0.0,
|
| 81 |
+
"eval_map_scissors": 0.0,
|
| 82 |
+
"eval_map_sheep": 0.0,
|
| 83 |
+
"eval_map_sink": 0.0001,
|
| 84 |
+
"eval_map_skateboard": 0.0,
|
| 85 |
+
"eval_map_skis": 0.0049,
|
| 86 |
+
"eval_map_small": 0.0085,
|
| 87 |
+
"eval_map_snowboard": 0.0,
|
| 88 |
+
"eval_map_spoon": 0.0,
|
| 89 |
+
"eval_map_sports ball": 0.0,
|
| 90 |
+
"eval_map_stop sign": 0.006,
|
| 91 |
+
"eval_map_suitcase": 0.0009,
|
| 92 |
+
"eval_map_surfboard": 0.0027,
|
| 93 |
+
"eval_map_teddy bear": 0.0,
|
| 94 |
+
"eval_map_tennis racket": 0.1978,
|
| 95 |
+
"eval_map_tie": 0.0,
|
| 96 |
+
"eval_map_toaster": -1.0,
|
| 97 |
+
"eval_map_toilet": 0.0101,
|
| 98 |
+
"eval_map_toothbrush": 0.0,
|
| 99 |
+
"eval_map_traffic light": 0.0388,
|
| 100 |
+
"eval_map_train": 0.0096,
|
| 101 |
+
"eval_map_truck": 0.108,
|
| 102 |
+
"eval_map_tv": 0.2457,
|
| 103 |
+
"eval_map_umbrella": 0.0001,
|
| 104 |
+
"eval_map_vase": 0.0002,
|
| 105 |
+
"eval_map_wine glass": 0.0,
|
| 106 |
+
"eval_map_zebra": 0.0015,
|
| 107 |
+
"eval_mar_1": 0.1037,
|
| 108 |
+
"eval_mar_10": 0.1704,
|
| 109 |
+
"eval_mar_100": 0.1887,
|
| 110 |
+
"eval_mar_100_airplane": 0.6857,
|
| 111 |
+
"eval_mar_100_apple": 0.0,
|
| 112 |
+
"eval_mar_100_backpack": 0.0333,
|
| 113 |
+
"eval_mar_100_banana": 0.0545,
|
| 114 |
+
"eval_mar_100_baseball bat": 0.0,
|
| 115 |
+
"eval_mar_100_baseball glove": 0.3438,
|
| 116 |
+
"eval_mar_100_bear": 0.0,
|
| 117 |
+
"eval_mar_100_bed": 0.0,
|
| 118 |
+
"eval_mar_100_bench": 0.275,
|
| 119 |
+
"eval_mar_100_bicycle": 0.0,
|
| 120 |
+
"eval_mar_100_bird": 0.5,
|
| 121 |
+
"eval_mar_100_boat": 0.2408,
|
| 122 |
+
"eval_mar_100_book": 0.0344,
|
| 123 |
+
"eval_mar_100_bottle": 0.3444,
|
| 124 |
+
"eval_mar_100_bowl": 0.224,
|
| 125 |
+
"eval_mar_100_broccoli": 0.1143,
|
| 126 |
+
"eval_mar_100_bus": 0.7917,
|
| 127 |
+
"eval_mar_100_cake": 0.25,
|
| 128 |
+
"eval_mar_100_car": 0.3451,
|
| 129 |
+
"eval_mar_100_carrot": 0.0,
|
| 130 |
+
"eval_mar_100_cat": 0.3692,
|
| 131 |
+
"eval_mar_100_cell phone": 0.1474,
|
| 132 |
+
"eval_mar_100_chair": 0.2458,
|
| 133 |
+
"eval_mar_100_clock": 0.1412,
|
| 134 |
+
"eval_mar_100_couch": 0.3455,
|
| 135 |
+
"eval_mar_100_cow": 0.4625,
|
| 136 |
+
"eval_mar_100_cup": 0.2686,
|
| 137 |
+
"eval_mar_100_dining table": 0.45,
|
| 138 |
+
"eval_mar_100_dog": 0.0,
|
| 139 |
+
"eval_mar_100_donut": -1.0,
|
| 140 |
+
"eval_mar_100_elephant": 0.5111,
|
| 141 |
+
"eval_mar_100_fire hydrant": -1.0,
|
| 142 |
+
"eval_mar_100_fork": 0.0643,
|
| 143 |
+
"eval_mar_100_frisbee": 0.6333,
|
| 144 |
+
"eval_mar_100_giraffe": 0.9556,
|
| 145 |
+
"eval_mar_100_handbag": 0.136,
|
| 146 |
+
"eval_mar_100_horse": 0.6,
|
| 147 |
+
"eval_mar_100_hot dog": 0.0,
|
| 148 |
+
"eval_mar_100_keyboard": 0.0,
|
| 149 |
+
"eval_mar_100_kite": 0.1611,
|
| 150 |
+
"eval_mar_100_knife": 0.0333,
|
| 151 |
+
"eval_mar_100_laptop": 0.1222,
|
| 152 |
+
"eval_mar_100_microwave": 0.0,
|
| 153 |
+
"eval_mar_100_motorcycle": 0.1437,
|
| 154 |
+
"eval_mar_100_mouse": 0.0,
|
| 155 |
+
"eval_mar_100_orange": 0.0,
|
| 156 |
+
"eval_mar_100_oven": 0.0,
|
| 157 |
+
"eval_mar_100_parking meter": 0.0,
|
| 158 |
+
"eval_mar_100_person": 0.4451,
|
| 159 |
+
"eval_mar_100_pizza": 0.16,
|
| 160 |
+
"eval_mar_100_potted plant": 0.205,
|
| 161 |
+
"eval_mar_100_refrigerator": 0.1286,
|
| 162 |
+
"eval_mar_100_remote": 0.0905,
|
| 163 |
+
"eval_mar_100_sandwich": 0.0,
|
| 164 |
+
"eval_mar_100_scissors": 0.0,
|
| 165 |
+
"eval_mar_100_sheep": 0.0,
|
| 166 |
+
"eval_mar_100_sink": 0.05,
|
| 167 |
+
"eval_mar_100_skateboard": 0.0,
|
| 168 |
+
"eval_mar_100_skis": 0.0875,
|
| 169 |
+
"eval_mar_100_snowboard": 0.0,
|
| 170 |
+
"eval_mar_100_spoon": 0.0,
|
| 171 |
+
"eval_mar_100_sports ball": 0.0,
|
| 172 |
+
"eval_mar_100_stop sign": 0.225,
|
| 173 |
+
"eval_mar_100_suitcase": 0.225,
|
| 174 |
+
"eval_mar_100_surfboard": 0.3818,
|
| 175 |
+
"eval_mar_100_teddy bear": 0.0,
|
| 176 |
+
"eval_mar_100_tennis racket": 0.5,
|
| 177 |
+
"eval_mar_100_tie": 0.0,
|
| 178 |
+
"eval_mar_100_toaster": -1.0,
|
| 179 |
+
"eval_mar_100_toilet": 0.2529,
|
| 180 |
+
"eval_mar_100_toothbrush": 0.0,
|
| 181 |
+
"eval_mar_100_traffic light": 0.1083,
|
| 182 |
+
"eval_mar_100_train": 0.1111,
|
| 183 |
+
"eval_mar_100_truck": 0.569,
|
| 184 |
+
"eval_mar_100_tv": 0.5615,
|
| 185 |
+
"eval_mar_100_umbrella": 0.0353,
|
| 186 |
+
"eval_mar_100_vase": 0.0833,
|
| 187 |
+
"eval_mar_100_wine glass": 0.0,
|
| 188 |
+
"eval_mar_100_zebra": 0.0909,
|
| 189 |
+
"eval_mar_large": 0.2753,
|
| 190 |
+
"eval_mar_medium": 0.1135,
|
| 191 |
+
"eval_mar_small": 0.0232,
|
| 192 |
+
"eval_runtime": 33.4497,
|
| 193 |
+
"eval_samples_per_second": 7.474,
|
| 194 |
+
"eval_steps_per_second": 7.474,
|
| 195 |
+
"step": 740
|
| 196 |
+
}
|
| 197 |
+
],
|
| 198 |
+
"logging_steps": 500,
|
| 199 |
+
"max_steps": 740,
|
| 200 |
+
"num_input_tokens_seen": 0,
|
| 201 |
+
"num_train_epochs": 1,
|
| 202 |
+
"save_steps": 500,
|
| 203 |
+
"stateful_callbacks": {
|
| 204 |
+
"TrainerControl": {
|
| 205 |
+
"args": {
|
| 206 |
+
"should_epoch_stop": false,
|
| 207 |
+
"should_evaluate": false,
|
| 208 |
+
"should_log": false,
|
| 209 |
+
"should_save": true,
|
| 210 |
+
"should_training_stop": true
|
| 211 |
+
},
|
| 212 |
+
"attributes": {}
|
| 213 |
+
}
|
| 214 |
+
},
|
| 215 |
+
"total_flos": 2.3008107254800712e+18,
|
| 216 |
+
"train_batch_size": 8,
|
| 217 |
+
"trial_name": null,
|
| 218 |
+
"trial_params": null
|
| 219 |
+
}
|
checkpoint-740/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:73718fb6f2ee2b9edbe5b3257b0d709323ef2b5e77752e6714b1bfd3ca3e7ab3
|
| 3 |
+
size 5841
|