Upload folder using huggingface_hub
Browse files- checkpoint-1021/config.json +284 -0
- checkpoint-1021/model.safetensors +3 -0
- checkpoint-1021/optimizer.pt +3 -0
- checkpoint-1021/preprocessor_config.json +24 -0
- checkpoint-1021/rng_state.pth +3 -0
- checkpoint-1021/scheduler.pt +3 -0
- checkpoint-1021/trainer_state.json +58 -0
- checkpoint-1021/training_args.bin +3 -0
- checkpoint-2042/config.json +284 -0
- checkpoint-2042/model.safetensors +3 -0
- checkpoint-2042/optimizer.pt +3 -0
- checkpoint-2042/preprocessor_config.json +24 -0
- checkpoint-2042/rng_state.pth +3 -0
- checkpoint-2042/scheduler.pt +3 -0
- checkpoint-2042/trainer_state.json +82 -0
- checkpoint-2042/training_args.bin +3 -0
- checkpoint-3063/config.json +284 -0
- checkpoint-3063/model.safetensors +3 -0
- checkpoint-3063/optimizer.pt +3 -0
- checkpoint-3063/preprocessor_config.json +24 -0
- checkpoint-3063/rng_state.pth +3 -0
- checkpoint-3063/scheduler.pt +3 -0
- checkpoint-3063/trainer_state.json +106 -0
- checkpoint-3063/training_args.bin +3 -0
- checkpoint-4084/config.json +284 -0
- checkpoint-4084/model.safetensors +3 -0
- checkpoint-4084/optimizer.pt +3 -0
- checkpoint-4084/preprocessor_config.json +24 -0
- checkpoint-4084/rng_state.pth +3 -0
- checkpoint-4084/scheduler.pt +3 -0
- checkpoint-4084/trainer_state.json +130 -0
- checkpoint-4084/training_args.bin +3 -0
- config.json +284 -0
- model.safetensors +3 -0
- preprocessor_config.json +24 -0
- training_args.bin +3 -0
checkpoint-1021/config.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "barn",
|
| 7 |
+
"1": "baseball_bat",
|
| 8 |
+
"2": "basket",
|
| 9 |
+
"3": "beach",
|
| 10 |
+
"4": "bear",
|
| 11 |
+
"5": "beard",
|
| 12 |
+
"6": "bee",
|
| 13 |
+
"7": "bird",
|
| 14 |
+
"8": "blueberry",
|
| 15 |
+
"9": "bowtie",
|
| 16 |
+
"10": "bracelet",
|
| 17 |
+
"11": "brain",
|
| 18 |
+
"12": "bread",
|
| 19 |
+
"13": "broccoli",
|
| 20 |
+
"14": "bus",
|
| 21 |
+
"15": "butterfly",
|
| 22 |
+
"16": "circle",
|
| 23 |
+
"17": "cloud",
|
| 24 |
+
"18": "cruise_ship",
|
| 25 |
+
"19": "dolphin",
|
| 26 |
+
"20": "dumbbell",
|
| 27 |
+
"21": "elephant",
|
| 28 |
+
"22": "eye",
|
| 29 |
+
"23": "eyeglasses",
|
| 30 |
+
"24": "feather",
|
| 31 |
+
"25": "fish",
|
| 32 |
+
"26": "flower",
|
| 33 |
+
"27": "foot",
|
| 34 |
+
"28": "frog",
|
| 35 |
+
"29": "giraffe",
|
| 36 |
+
"30": "goatee",
|
| 37 |
+
"31": "golf_club",
|
| 38 |
+
"32": "grapes",
|
| 39 |
+
"33": "grass",
|
| 40 |
+
"34": "guitar",
|
| 41 |
+
"35": "hamburger",
|
| 42 |
+
"36": "hand",
|
| 43 |
+
"37": "hat",
|
| 44 |
+
"38": "headphones",
|
| 45 |
+
"39": "helicopter",
|
| 46 |
+
"40": "hexagon",
|
| 47 |
+
"41": "hockey_stick",
|
| 48 |
+
"42": "horse",
|
| 49 |
+
"43": "hourglass",
|
| 50 |
+
"44": "house",
|
| 51 |
+
"45": "ice_cream",
|
| 52 |
+
"46": "jacket",
|
| 53 |
+
"47": "ladder",
|
| 54 |
+
"48": "leg",
|
| 55 |
+
"49": "lipstick",
|
| 56 |
+
"50": "megaphone",
|
| 57 |
+
"51": "monkey",
|
| 58 |
+
"52": "moon",
|
| 59 |
+
"53": "mushroom",
|
| 60 |
+
"54": "necklace",
|
| 61 |
+
"55": "owl",
|
| 62 |
+
"56": "panda",
|
| 63 |
+
"57": "pear",
|
| 64 |
+
"58": "peas",
|
| 65 |
+
"59": "penguin",
|
| 66 |
+
"60": "pig",
|
| 67 |
+
"61": "pillow",
|
| 68 |
+
"62": "pineapple",
|
| 69 |
+
"63": "pizza",
|
| 70 |
+
"64": "pool",
|
| 71 |
+
"65": "popsicle",
|
| 72 |
+
"66": "rabbit",
|
| 73 |
+
"67": "rhinoceros",
|
| 74 |
+
"68": "rifle",
|
| 75 |
+
"69": "river",
|
| 76 |
+
"70": "sailboat",
|
| 77 |
+
"71": "sandwich",
|
| 78 |
+
"72": "sea_turtle",
|
| 79 |
+
"73": "shark",
|
| 80 |
+
"74": "shoe",
|
| 81 |
+
"75": "skyscraper",
|
| 82 |
+
"76": "snorkel",
|
| 83 |
+
"77": "snowman",
|
| 84 |
+
"78": "soccer_ball",
|
| 85 |
+
"79": "speedboat",
|
| 86 |
+
"80": "spider",
|
| 87 |
+
"81": "spoon",
|
| 88 |
+
"82": "square",
|
| 89 |
+
"83": "squirrel",
|
| 90 |
+
"84": "stethoscope",
|
| 91 |
+
"85": "strawberry",
|
| 92 |
+
"86": "streetlight",
|
| 93 |
+
"87": "submarine",
|
| 94 |
+
"88": "suitcase",
|
| 95 |
+
"89": "sun",
|
| 96 |
+
"90": "sweater",
|
| 97 |
+
"91": "sword",
|
| 98 |
+
"92": "table",
|
| 99 |
+
"93": "teapot",
|
| 100 |
+
"94": "teddy-bear",
|
| 101 |
+
"95": "telephone",
|
| 102 |
+
"96": "tent",
|
| 103 |
+
"97": "The_Eiffel_Tower",
|
| 104 |
+
"98": "The_Great_Wall_of_China",
|
| 105 |
+
"99": "The_Mona_Lisa",
|
| 106 |
+
"100": "tiger",
|
| 107 |
+
"101": "toaster",
|
| 108 |
+
"102": "tooth",
|
| 109 |
+
"103": "tornado",
|
| 110 |
+
"104": "tractor",
|
| 111 |
+
"105": "train",
|
| 112 |
+
"106": "tree",
|
| 113 |
+
"107": "triangle",
|
| 114 |
+
"108": "trombone",
|
| 115 |
+
"109": "truck",
|
| 116 |
+
"110": "trumpet",
|
| 117 |
+
"111": "umbrella",
|
| 118 |
+
"112": "vase",
|
| 119 |
+
"113": "violin",
|
| 120 |
+
"114": "watermelon",
|
| 121 |
+
"115": "whale",
|
| 122 |
+
"116": "windmill",
|
| 123 |
+
"117": "wine_glass",
|
| 124 |
+
"118": "yoga",
|
| 125 |
+
"119": "zebra",
|
| 126 |
+
"120": "zigzag"
|
| 127 |
+
},
|
| 128 |
+
"initializer_factor": 1.0,
|
| 129 |
+
"label2id": {
|
| 130 |
+
"The_Eiffel_Tower": 97,
|
| 131 |
+
"The_Great_Wall_of_China": 98,
|
| 132 |
+
"The_Mona_Lisa": 99,
|
| 133 |
+
"barn": 0,
|
| 134 |
+
"baseball_bat": 1,
|
| 135 |
+
"basket": 2,
|
| 136 |
+
"beach": 3,
|
| 137 |
+
"bear": 4,
|
| 138 |
+
"beard": 5,
|
| 139 |
+
"bee": 6,
|
| 140 |
+
"bird": 7,
|
| 141 |
+
"blueberry": 8,
|
| 142 |
+
"bowtie": 9,
|
| 143 |
+
"bracelet": 10,
|
| 144 |
+
"brain": 11,
|
| 145 |
+
"bread": 12,
|
| 146 |
+
"broccoli": 13,
|
| 147 |
+
"bus": 14,
|
| 148 |
+
"butterfly": 15,
|
| 149 |
+
"circle": 16,
|
| 150 |
+
"cloud": 17,
|
| 151 |
+
"cruise_ship": 18,
|
| 152 |
+
"dolphin": 19,
|
| 153 |
+
"dumbbell": 20,
|
| 154 |
+
"elephant": 21,
|
| 155 |
+
"eye": 22,
|
| 156 |
+
"eyeglasses": 23,
|
| 157 |
+
"feather": 24,
|
| 158 |
+
"fish": 25,
|
| 159 |
+
"flower": 26,
|
| 160 |
+
"foot": 27,
|
| 161 |
+
"frog": 28,
|
| 162 |
+
"giraffe": 29,
|
| 163 |
+
"goatee": 30,
|
| 164 |
+
"golf_club": 31,
|
| 165 |
+
"grapes": 32,
|
| 166 |
+
"grass": 33,
|
| 167 |
+
"guitar": 34,
|
| 168 |
+
"hamburger": 35,
|
| 169 |
+
"hand": 36,
|
| 170 |
+
"hat": 37,
|
| 171 |
+
"headphones": 38,
|
| 172 |
+
"helicopter": 39,
|
| 173 |
+
"hexagon": 40,
|
| 174 |
+
"hockey_stick": 41,
|
| 175 |
+
"horse": 42,
|
| 176 |
+
"hourglass": 43,
|
| 177 |
+
"house": 44,
|
| 178 |
+
"ice_cream": 45,
|
| 179 |
+
"jacket": 46,
|
| 180 |
+
"ladder": 47,
|
| 181 |
+
"leg": 48,
|
| 182 |
+
"lipstick": 49,
|
| 183 |
+
"megaphone": 50,
|
| 184 |
+
"monkey": 51,
|
| 185 |
+
"moon": 52,
|
| 186 |
+
"mushroom": 53,
|
| 187 |
+
"necklace": 54,
|
| 188 |
+
"owl": 55,
|
| 189 |
+
"panda": 56,
|
| 190 |
+
"pear": 57,
|
| 191 |
+
"peas": 58,
|
| 192 |
+
"penguin": 59,
|
| 193 |
+
"pig": 60,
|
| 194 |
+
"pillow": 61,
|
| 195 |
+
"pineapple": 62,
|
| 196 |
+
"pizza": 63,
|
| 197 |
+
"pool": 64,
|
| 198 |
+
"popsicle": 65,
|
| 199 |
+
"rabbit": 66,
|
| 200 |
+
"rhinoceros": 67,
|
| 201 |
+
"rifle": 68,
|
| 202 |
+
"river": 69,
|
| 203 |
+
"sailboat": 70,
|
| 204 |
+
"sandwich": 71,
|
| 205 |
+
"sea_turtle": 72,
|
| 206 |
+
"shark": 73,
|
| 207 |
+
"shoe": 74,
|
| 208 |
+
"skyscraper": 75,
|
| 209 |
+
"snorkel": 76,
|
| 210 |
+
"snowman": 77,
|
| 211 |
+
"soccer_ball": 78,
|
| 212 |
+
"speedboat": 79,
|
| 213 |
+
"spider": 80,
|
| 214 |
+
"spoon": 81,
|
| 215 |
+
"square": 82,
|
| 216 |
+
"squirrel": 83,
|
| 217 |
+
"stethoscope": 84,
|
| 218 |
+
"strawberry": 85,
|
| 219 |
+
"streetlight": 86,
|
| 220 |
+
"submarine": 87,
|
| 221 |
+
"suitcase": 88,
|
| 222 |
+
"sun": 89,
|
| 223 |
+
"sweater": 90,
|
| 224 |
+
"sword": 91,
|
| 225 |
+
"table": 92,
|
| 226 |
+
"teapot": 93,
|
| 227 |
+
"teddy-bear": 94,
|
| 228 |
+
"telephone": 95,
|
| 229 |
+
"tent": 96,
|
| 230 |
+
"tiger": 100,
|
| 231 |
+
"toaster": 101,
|
| 232 |
+
"tooth": 102,
|
| 233 |
+
"tornado": 103,
|
| 234 |
+
"tractor": 104,
|
| 235 |
+
"train": 105,
|
| 236 |
+
"tree": 106,
|
| 237 |
+
"triangle": 107,
|
| 238 |
+
"trombone": 108,
|
| 239 |
+
"truck": 109,
|
| 240 |
+
"trumpet": 110,
|
| 241 |
+
"umbrella": 111,
|
| 242 |
+
"vase": 112,
|
| 243 |
+
"violin": 113,
|
| 244 |
+
"watermelon": 114,
|
| 245 |
+
"whale": 115,
|
| 246 |
+
"windmill": 116,
|
| 247 |
+
"wine_glass": 117,
|
| 248 |
+
"yoga": 118,
|
| 249 |
+
"zebra": 119,
|
| 250 |
+
"zigzag": 120
|
| 251 |
+
},
|
| 252 |
+
"model_type": "siglip",
|
| 253 |
+
"problem_type": "single_label_classification",
|
| 254 |
+
"text_config": {
|
| 255 |
+
"attention_dropout": 0.0,
|
| 256 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 257 |
+
"hidden_size": 768,
|
| 258 |
+
"intermediate_size": 3072,
|
| 259 |
+
"layer_norm_eps": 1e-06,
|
| 260 |
+
"max_position_embeddings": 64,
|
| 261 |
+
"model_type": "siglip_text_model",
|
| 262 |
+
"num_attention_heads": 12,
|
| 263 |
+
"num_hidden_layers": 12,
|
| 264 |
+
"projection_size": 768,
|
| 265 |
+
"torch_dtype": "float32",
|
| 266 |
+
"vocab_size": 256000
|
| 267 |
+
},
|
| 268 |
+
"torch_dtype": "float32",
|
| 269 |
+
"transformers_version": "4.51.0.dev0",
|
| 270 |
+
"vision_config": {
|
| 271 |
+
"attention_dropout": 0.0,
|
| 272 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 273 |
+
"hidden_size": 768,
|
| 274 |
+
"image_size": 224,
|
| 275 |
+
"intermediate_size": 3072,
|
| 276 |
+
"layer_norm_eps": 1e-06,
|
| 277 |
+
"model_type": "siglip_vision_model",
|
| 278 |
+
"num_attention_heads": 12,
|
| 279 |
+
"num_channels": 3,
|
| 280 |
+
"num_hidden_layers": 12,
|
| 281 |
+
"patch_size": 16,
|
| 282 |
+
"torch_dtype": "float32"
|
| 283 |
+
}
|
| 284 |
+
}
|
checkpoint-1021/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:707146d57207c7e741dde1985e283cd0c9b2e6f94c67fd4afc74362bb59f9a29
|
| 3 |
+
size 371934052
|
checkpoint-1021/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d9cfa7be0915bddb120a1676cbf1b331993bcb3f9404f2e428f1099b7d570f8e
|
| 3 |
+
size 687287802
|
checkpoint-1021/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-1021/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6cf88982be55987cb2658acaf24ac8215ddd09c1d938d19260cd24a48bf7d455
|
| 3 |
+
size 14244
|
checkpoint-1021/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cde7e9442d417bc65f8bac8f1267e42e2bd5cbcb60c34f92926e7ff3a3b92a3d
|
| 3 |
+
size 1064
|
checkpoint-1021/trainer_state.json
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 1021,
|
| 3 |
+
"best_metric": 1.7632426023483276,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-1021",
|
| 5 |
+
"epoch": 1.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 1021,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.48971596474045054,
|
| 14 |
+
"grad_norm": 59.27558135986328,
|
| 15 |
+
"learning_rate": 2.665344571145265e-06,
|
| 16 |
+
"loss": 4.268,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.9794319294809011,
|
| 21 |
+
"grad_norm": 61.66608810424805,
|
| 22 |
+
"learning_rate": 2.2935052057511156e-06,
|
| 23 |
+
"loss": 2.4217,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.666574839302112,
|
| 29 |
+
"eval_loss": 1.7632426023483276,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 476.9192,
|
| 32 |
+
"eval_samples_per_second": 68.502,
|
| 33 |
+
"eval_steps_per_second": 8.563,
|
| 34 |
+
"step": 1021
|
| 35 |
+
}
|
| 36 |
+
],
|
| 37 |
+
"logging_steps": 500,
|
| 38 |
+
"max_steps": 4084,
|
| 39 |
+
"num_input_tokens_seen": 0,
|
| 40 |
+
"num_train_epochs": 4,
|
| 41 |
+
"save_steps": 500,
|
| 42 |
+
"stateful_callbacks": {
|
| 43 |
+
"TrainerControl": {
|
| 44 |
+
"args": {
|
| 45 |
+
"should_epoch_stop": false,
|
| 46 |
+
"should_evaluate": false,
|
| 47 |
+
"should_log": false,
|
| 48 |
+
"should_save": true,
|
| 49 |
+
"should_training_stop": false
|
| 50 |
+
},
|
| 51 |
+
"attributes": {}
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"total_flos": 2.738992217671987e+18,
|
| 55 |
+
"train_batch_size": 32,
|
| 56 |
+
"trial_name": null,
|
| 57 |
+
"trial_params": null
|
| 58 |
+
}
|
checkpoint-1021/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eff7895602b654dfe6fc27f9bfd70976ed586448b1d172d1abb6cf2d3bccf43
|
| 3 |
+
size 5304
|
checkpoint-2042/config.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "barn",
|
| 7 |
+
"1": "baseball_bat",
|
| 8 |
+
"2": "basket",
|
| 9 |
+
"3": "beach",
|
| 10 |
+
"4": "bear",
|
| 11 |
+
"5": "beard",
|
| 12 |
+
"6": "bee",
|
| 13 |
+
"7": "bird",
|
| 14 |
+
"8": "blueberry",
|
| 15 |
+
"9": "bowtie",
|
| 16 |
+
"10": "bracelet",
|
| 17 |
+
"11": "brain",
|
| 18 |
+
"12": "bread",
|
| 19 |
+
"13": "broccoli",
|
| 20 |
+
"14": "bus",
|
| 21 |
+
"15": "butterfly",
|
| 22 |
+
"16": "circle",
|
| 23 |
+
"17": "cloud",
|
| 24 |
+
"18": "cruise_ship",
|
| 25 |
+
"19": "dolphin",
|
| 26 |
+
"20": "dumbbell",
|
| 27 |
+
"21": "elephant",
|
| 28 |
+
"22": "eye",
|
| 29 |
+
"23": "eyeglasses",
|
| 30 |
+
"24": "feather",
|
| 31 |
+
"25": "fish",
|
| 32 |
+
"26": "flower",
|
| 33 |
+
"27": "foot",
|
| 34 |
+
"28": "frog",
|
| 35 |
+
"29": "giraffe",
|
| 36 |
+
"30": "goatee",
|
| 37 |
+
"31": "golf_club",
|
| 38 |
+
"32": "grapes",
|
| 39 |
+
"33": "grass",
|
| 40 |
+
"34": "guitar",
|
| 41 |
+
"35": "hamburger",
|
| 42 |
+
"36": "hand",
|
| 43 |
+
"37": "hat",
|
| 44 |
+
"38": "headphones",
|
| 45 |
+
"39": "helicopter",
|
| 46 |
+
"40": "hexagon",
|
| 47 |
+
"41": "hockey_stick",
|
| 48 |
+
"42": "horse",
|
| 49 |
+
"43": "hourglass",
|
| 50 |
+
"44": "house",
|
| 51 |
+
"45": "ice_cream",
|
| 52 |
+
"46": "jacket",
|
| 53 |
+
"47": "ladder",
|
| 54 |
+
"48": "leg",
|
| 55 |
+
"49": "lipstick",
|
| 56 |
+
"50": "megaphone",
|
| 57 |
+
"51": "monkey",
|
| 58 |
+
"52": "moon",
|
| 59 |
+
"53": "mushroom",
|
| 60 |
+
"54": "necklace",
|
| 61 |
+
"55": "owl",
|
| 62 |
+
"56": "panda",
|
| 63 |
+
"57": "pear",
|
| 64 |
+
"58": "peas",
|
| 65 |
+
"59": "penguin",
|
| 66 |
+
"60": "pig",
|
| 67 |
+
"61": "pillow",
|
| 68 |
+
"62": "pineapple",
|
| 69 |
+
"63": "pizza",
|
| 70 |
+
"64": "pool",
|
| 71 |
+
"65": "popsicle",
|
| 72 |
+
"66": "rabbit",
|
| 73 |
+
"67": "rhinoceros",
|
| 74 |
+
"68": "rifle",
|
| 75 |
+
"69": "river",
|
| 76 |
+
"70": "sailboat",
|
| 77 |
+
"71": "sandwich",
|
| 78 |
+
"72": "sea_turtle",
|
| 79 |
+
"73": "shark",
|
| 80 |
+
"74": "shoe",
|
| 81 |
+
"75": "skyscraper",
|
| 82 |
+
"76": "snorkel",
|
| 83 |
+
"77": "snowman",
|
| 84 |
+
"78": "soccer_ball",
|
| 85 |
+
"79": "speedboat",
|
| 86 |
+
"80": "spider",
|
| 87 |
+
"81": "spoon",
|
| 88 |
+
"82": "square",
|
| 89 |
+
"83": "squirrel",
|
| 90 |
+
"84": "stethoscope",
|
| 91 |
+
"85": "strawberry",
|
| 92 |
+
"86": "streetlight",
|
| 93 |
+
"87": "submarine",
|
| 94 |
+
"88": "suitcase",
|
| 95 |
+
"89": "sun",
|
| 96 |
+
"90": "sweater",
|
| 97 |
+
"91": "sword",
|
| 98 |
+
"92": "table",
|
| 99 |
+
"93": "teapot",
|
| 100 |
+
"94": "teddy-bear",
|
| 101 |
+
"95": "telephone",
|
| 102 |
+
"96": "tent",
|
| 103 |
+
"97": "The_Eiffel_Tower",
|
| 104 |
+
"98": "The_Great_Wall_of_China",
|
| 105 |
+
"99": "The_Mona_Lisa",
|
| 106 |
+
"100": "tiger",
|
| 107 |
+
"101": "toaster",
|
| 108 |
+
"102": "tooth",
|
| 109 |
+
"103": "tornado",
|
| 110 |
+
"104": "tractor",
|
| 111 |
+
"105": "train",
|
| 112 |
+
"106": "tree",
|
| 113 |
+
"107": "triangle",
|
| 114 |
+
"108": "trombone",
|
| 115 |
+
"109": "truck",
|
| 116 |
+
"110": "trumpet",
|
| 117 |
+
"111": "umbrella",
|
| 118 |
+
"112": "vase",
|
| 119 |
+
"113": "violin",
|
| 120 |
+
"114": "watermelon",
|
| 121 |
+
"115": "whale",
|
| 122 |
+
"116": "windmill",
|
| 123 |
+
"117": "wine_glass",
|
| 124 |
+
"118": "yoga",
|
| 125 |
+
"119": "zebra",
|
| 126 |
+
"120": "zigzag"
|
| 127 |
+
},
|
| 128 |
+
"initializer_factor": 1.0,
|
| 129 |
+
"label2id": {
|
| 130 |
+
"The_Eiffel_Tower": 97,
|
| 131 |
+
"The_Great_Wall_of_China": 98,
|
| 132 |
+
"The_Mona_Lisa": 99,
|
| 133 |
+
"barn": 0,
|
| 134 |
+
"baseball_bat": 1,
|
| 135 |
+
"basket": 2,
|
| 136 |
+
"beach": 3,
|
| 137 |
+
"bear": 4,
|
| 138 |
+
"beard": 5,
|
| 139 |
+
"bee": 6,
|
| 140 |
+
"bird": 7,
|
| 141 |
+
"blueberry": 8,
|
| 142 |
+
"bowtie": 9,
|
| 143 |
+
"bracelet": 10,
|
| 144 |
+
"brain": 11,
|
| 145 |
+
"bread": 12,
|
| 146 |
+
"broccoli": 13,
|
| 147 |
+
"bus": 14,
|
| 148 |
+
"butterfly": 15,
|
| 149 |
+
"circle": 16,
|
| 150 |
+
"cloud": 17,
|
| 151 |
+
"cruise_ship": 18,
|
| 152 |
+
"dolphin": 19,
|
| 153 |
+
"dumbbell": 20,
|
| 154 |
+
"elephant": 21,
|
| 155 |
+
"eye": 22,
|
| 156 |
+
"eyeglasses": 23,
|
| 157 |
+
"feather": 24,
|
| 158 |
+
"fish": 25,
|
| 159 |
+
"flower": 26,
|
| 160 |
+
"foot": 27,
|
| 161 |
+
"frog": 28,
|
| 162 |
+
"giraffe": 29,
|
| 163 |
+
"goatee": 30,
|
| 164 |
+
"golf_club": 31,
|
| 165 |
+
"grapes": 32,
|
| 166 |
+
"grass": 33,
|
| 167 |
+
"guitar": 34,
|
| 168 |
+
"hamburger": 35,
|
| 169 |
+
"hand": 36,
|
| 170 |
+
"hat": 37,
|
| 171 |
+
"headphones": 38,
|
| 172 |
+
"helicopter": 39,
|
| 173 |
+
"hexagon": 40,
|
| 174 |
+
"hockey_stick": 41,
|
| 175 |
+
"horse": 42,
|
| 176 |
+
"hourglass": 43,
|
| 177 |
+
"house": 44,
|
| 178 |
+
"ice_cream": 45,
|
| 179 |
+
"jacket": 46,
|
| 180 |
+
"ladder": 47,
|
| 181 |
+
"leg": 48,
|
| 182 |
+
"lipstick": 49,
|
| 183 |
+
"megaphone": 50,
|
| 184 |
+
"monkey": 51,
|
| 185 |
+
"moon": 52,
|
| 186 |
+
"mushroom": 53,
|
| 187 |
+
"necklace": 54,
|
| 188 |
+
"owl": 55,
|
| 189 |
+
"panda": 56,
|
| 190 |
+
"pear": 57,
|
| 191 |
+
"peas": 58,
|
| 192 |
+
"penguin": 59,
|
| 193 |
+
"pig": 60,
|
| 194 |
+
"pillow": 61,
|
| 195 |
+
"pineapple": 62,
|
| 196 |
+
"pizza": 63,
|
| 197 |
+
"pool": 64,
|
| 198 |
+
"popsicle": 65,
|
| 199 |
+
"rabbit": 66,
|
| 200 |
+
"rhinoceros": 67,
|
| 201 |
+
"rifle": 68,
|
| 202 |
+
"river": 69,
|
| 203 |
+
"sailboat": 70,
|
| 204 |
+
"sandwich": 71,
|
| 205 |
+
"sea_turtle": 72,
|
| 206 |
+
"shark": 73,
|
| 207 |
+
"shoe": 74,
|
| 208 |
+
"skyscraper": 75,
|
| 209 |
+
"snorkel": 76,
|
| 210 |
+
"snowman": 77,
|
| 211 |
+
"soccer_ball": 78,
|
| 212 |
+
"speedboat": 79,
|
| 213 |
+
"spider": 80,
|
| 214 |
+
"spoon": 81,
|
| 215 |
+
"square": 82,
|
| 216 |
+
"squirrel": 83,
|
| 217 |
+
"stethoscope": 84,
|
| 218 |
+
"strawberry": 85,
|
| 219 |
+
"streetlight": 86,
|
| 220 |
+
"submarine": 87,
|
| 221 |
+
"suitcase": 88,
|
| 222 |
+
"sun": 89,
|
| 223 |
+
"sweater": 90,
|
| 224 |
+
"sword": 91,
|
| 225 |
+
"table": 92,
|
| 226 |
+
"teapot": 93,
|
| 227 |
+
"teddy-bear": 94,
|
| 228 |
+
"telephone": 95,
|
| 229 |
+
"tent": 96,
|
| 230 |
+
"tiger": 100,
|
| 231 |
+
"toaster": 101,
|
| 232 |
+
"tooth": 102,
|
| 233 |
+
"tornado": 103,
|
| 234 |
+
"tractor": 104,
|
| 235 |
+
"train": 105,
|
| 236 |
+
"tree": 106,
|
| 237 |
+
"triangle": 107,
|
| 238 |
+
"trombone": 108,
|
| 239 |
+
"truck": 109,
|
| 240 |
+
"trumpet": 110,
|
| 241 |
+
"umbrella": 111,
|
| 242 |
+
"vase": 112,
|
| 243 |
+
"violin": 113,
|
| 244 |
+
"watermelon": 114,
|
| 245 |
+
"whale": 115,
|
| 246 |
+
"windmill": 116,
|
| 247 |
+
"wine_glass": 117,
|
| 248 |
+
"yoga": 118,
|
| 249 |
+
"zebra": 119,
|
| 250 |
+
"zigzag": 120
|
| 251 |
+
},
|
| 252 |
+
"model_type": "siglip",
|
| 253 |
+
"problem_type": "single_label_classification",
|
| 254 |
+
"text_config": {
|
| 255 |
+
"attention_dropout": 0.0,
|
| 256 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 257 |
+
"hidden_size": 768,
|
| 258 |
+
"intermediate_size": 3072,
|
| 259 |
+
"layer_norm_eps": 1e-06,
|
| 260 |
+
"max_position_embeddings": 64,
|
| 261 |
+
"model_type": "siglip_text_model",
|
| 262 |
+
"num_attention_heads": 12,
|
| 263 |
+
"num_hidden_layers": 12,
|
| 264 |
+
"projection_size": 768,
|
| 265 |
+
"torch_dtype": "float32",
|
| 266 |
+
"vocab_size": 256000
|
| 267 |
+
},
|
| 268 |
+
"torch_dtype": "float32",
|
| 269 |
+
"transformers_version": "4.51.0.dev0",
|
| 270 |
+
"vision_config": {
|
| 271 |
+
"attention_dropout": 0.0,
|
| 272 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 273 |
+
"hidden_size": 768,
|
| 274 |
+
"image_size": 224,
|
| 275 |
+
"intermediate_size": 3072,
|
| 276 |
+
"layer_norm_eps": 1e-06,
|
| 277 |
+
"model_type": "siglip_vision_model",
|
| 278 |
+
"num_attention_heads": 12,
|
| 279 |
+
"num_channels": 3,
|
| 280 |
+
"num_hidden_layers": 12,
|
| 281 |
+
"patch_size": 16,
|
| 282 |
+
"torch_dtype": "float32"
|
| 283 |
+
}
|
| 284 |
+
}
|
checkpoint-2042/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed04b715193ee61b7818057c38a6701aa9d73dd798f90d1ff53a590b50b5ab9e
|
| 3 |
+
size 371934052
|
checkpoint-2042/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:45359769ce363f07fdafc67a3880f22848e75a57d269b7e544eafe492a9f18b2
|
| 3 |
+
size 687287802
|
checkpoint-2042/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-2042/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d44c4a6883b1580c48141c1c6dea8913f0a639487556e60c2261d947215c3bcb
|
| 3 |
+
size 14244
|
checkpoint-2042/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:072e1205b368312b45b1d5485fd500d89e1a5154f58c57dbe12638c8bae74e82
|
| 3 |
+
size 1064
|
checkpoint-2042/trainer_state.json
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 2042,
|
| 3 |
+
"best_metric": 1.171113133430481,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-2042",
|
| 5 |
+
"epoch": 2.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 2042,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.48971596474045054,
|
| 14 |
+
"grad_norm": 59.27558135986328,
|
| 15 |
+
"learning_rate": 2.665344571145265e-06,
|
| 16 |
+
"loss": 4.268,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.9794319294809011,
|
| 21 |
+
"grad_norm": 61.66608810424805,
|
| 22 |
+
"learning_rate": 2.2935052057511156e-06,
|
| 23 |
+
"loss": 2.4217,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.666574839302112,
|
| 29 |
+
"eval_loss": 1.7632426023483276,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 476.9192,
|
| 32 |
+
"eval_samples_per_second": 68.502,
|
| 33 |
+
"eval_steps_per_second": 8.563,
|
| 34 |
+
"step": 1021
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 1.4691478942213516,
|
| 38 |
+
"grad_norm": 71.97106170654297,
|
| 39 |
+
"learning_rate": 1.921665840356966e-06,
|
| 40 |
+
"loss": 1.709,
|
| 41 |
+
"step": 1500
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 1.9588638589618022,
|
| 45 |
+
"grad_norm": 63.00515365600586,
|
| 46 |
+
"learning_rate": 1.5498264749628163e-06,
|
| 47 |
+
"loss": 1.4697,
|
| 48 |
+
"step": 2000
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 2.0,
|
| 52 |
+
"eval_accuracy": 0.7615243342516069,
|
| 53 |
+
"eval_loss": 1.171113133430481,
|
| 54 |
+
"eval_model_preparation_time": 0.0024,
|
| 55 |
+
"eval_runtime": 475.7496,
|
| 56 |
+
"eval_samples_per_second": 68.671,
|
| 57 |
+
"eval_steps_per_second": 8.584,
|
| 58 |
+
"step": 2042
|
| 59 |
+
}
|
| 60 |
+
],
|
| 61 |
+
"logging_steps": 500,
|
| 62 |
+
"max_steps": 4084,
|
| 63 |
+
"num_input_tokens_seen": 0,
|
| 64 |
+
"num_train_epochs": 4,
|
| 65 |
+
"save_steps": 500,
|
| 66 |
+
"stateful_callbacks": {
|
| 67 |
+
"TrainerControl": {
|
| 68 |
+
"args": {
|
| 69 |
+
"should_epoch_stop": false,
|
| 70 |
+
"should_evaluate": false,
|
| 71 |
+
"should_log": false,
|
| 72 |
+
"should_save": true,
|
| 73 |
+
"should_training_stop": false
|
| 74 |
+
},
|
| 75 |
+
"attributes": {}
|
| 76 |
+
}
|
| 77 |
+
},
|
| 78 |
+
"total_flos": 5.477984435343974e+18,
|
| 79 |
+
"train_batch_size": 32,
|
| 80 |
+
"trial_name": null,
|
| 81 |
+
"trial_params": null
|
| 82 |
+
}
|
checkpoint-2042/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eff7895602b654dfe6fc27f9bfd70976ed586448b1d172d1abb6cf2d3bccf43
|
| 3 |
+
size 5304
|
checkpoint-3063/config.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "barn",
|
| 7 |
+
"1": "baseball_bat",
|
| 8 |
+
"2": "basket",
|
| 9 |
+
"3": "beach",
|
| 10 |
+
"4": "bear",
|
| 11 |
+
"5": "beard",
|
| 12 |
+
"6": "bee",
|
| 13 |
+
"7": "bird",
|
| 14 |
+
"8": "blueberry",
|
| 15 |
+
"9": "bowtie",
|
| 16 |
+
"10": "bracelet",
|
| 17 |
+
"11": "brain",
|
| 18 |
+
"12": "bread",
|
| 19 |
+
"13": "broccoli",
|
| 20 |
+
"14": "bus",
|
| 21 |
+
"15": "butterfly",
|
| 22 |
+
"16": "circle",
|
| 23 |
+
"17": "cloud",
|
| 24 |
+
"18": "cruise_ship",
|
| 25 |
+
"19": "dolphin",
|
| 26 |
+
"20": "dumbbell",
|
| 27 |
+
"21": "elephant",
|
| 28 |
+
"22": "eye",
|
| 29 |
+
"23": "eyeglasses",
|
| 30 |
+
"24": "feather",
|
| 31 |
+
"25": "fish",
|
| 32 |
+
"26": "flower",
|
| 33 |
+
"27": "foot",
|
| 34 |
+
"28": "frog",
|
| 35 |
+
"29": "giraffe",
|
| 36 |
+
"30": "goatee",
|
| 37 |
+
"31": "golf_club",
|
| 38 |
+
"32": "grapes",
|
| 39 |
+
"33": "grass",
|
| 40 |
+
"34": "guitar",
|
| 41 |
+
"35": "hamburger",
|
| 42 |
+
"36": "hand",
|
| 43 |
+
"37": "hat",
|
| 44 |
+
"38": "headphones",
|
| 45 |
+
"39": "helicopter",
|
| 46 |
+
"40": "hexagon",
|
| 47 |
+
"41": "hockey_stick",
|
| 48 |
+
"42": "horse",
|
| 49 |
+
"43": "hourglass",
|
| 50 |
+
"44": "house",
|
| 51 |
+
"45": "ice_cream",
|
| 52 |
+
"46": "jacket",
|
| 53 |
+
"47": "ladder",
|
| 54 |
+
"48": "leg",
|
| 55 |
+
"49": "lipstick",
|
| 56 |
+
"50": "megaphone",
|
| 57 |
+
"51": "monkey",
|
| 58 |
+
"52": "moon",
|
| 59 |
+
"53": "mushroom",
|
| 60 |
+
"54": "necklace",
|
| 61 |
+
"55": "owl",
|
| 62 |
+
"56": "panda",
|
| 63 |
+
"57": "pear",
|
| 64 |
+
"58": "peas",
|
| 65 |
+
"59": "penguin",
|
| 66 |
+
"60": "pig",
|
| 67 |
+
"61": "pillow",
|
| 68 |
+
"62": "pineapple",
|
| 69 |
+
"63": "pizza",
|
| 70 |
+
"64": "pool",
|
| 71 |
+
"65": "popsicle",
|
| 72 |
+
"66": "rabbit",
|
| 73 |
+
"67": "rhinoceros",
|
| 74 |
+
"68": "rifle",
|
| 75 |
+
"69": "river",
|
| 76 |
+
"70": "sailboat",
|
| 77 |
+
"71": "sandwich",
|
| 78 |
+
"72": "sea_turtle",
|
| 79 |
+
"73": "shark",
|
| 80 |
+
"74": "shoe",
|
| 81 |
+
"75": "skyscraper",
|
| 82 |
+
"76": "snorkel",
|
| 83 |
+
"77": "snowman",
|
| 84 |
+
"78": "soccer_ball",
|
| 85 |
+
"79": "speedboat",
|
| 86 |
+
"80": "spider",
|
| 87 |
+
"81": "spoon",
|
| 88 |
+
"82": "square",
|
| 89 |
+
"83": "squirrel",
|
| 90 |
+
"84": "stethoscope",
|
| 91 |
+
"85": "strawberry",
|
| 92 |
+
"86": "streetlight",
|
| 93 |
+
"87": "submarine",
|
| 94 |
+
"88": "suitcase",
|
| 95 |
+
"89": "sun",
|
| 96 |
+
"90": "sweater",
|
| 97 |
+
"91": "sword",
|
| 98 |
+
"92": "table",
|
| 99 |
+
"93": "teapot",
|
| 100 |
+
"94": "teddy-bear",
|
| 101 |
+
"95": "telephone",
|
| 102 |
+
"96": "tent",
|
| 103 |
+
"97": "The_Eiffel_Tower",
|
| 104 |
+
"98": "The_Great_Wall_of_China",
|
| 105 |
+
"99": "The_Mona_Lisa",
|
| 106 |
+
"100": "tiger",
|
| 107 |
+
"101": "toaster",
|
| 108 |
+
"102": "tooth",
|
| 109 |
+
"103": "tornado",
|
| 110 |
+
"104": "tractor",
|
| 111 |
+
"105": "train",
|
| 112 |
+
"106": "tree",
|
| 113 |
+
"107": "triangle",
|
| 114 |
+
"108": "trombone",
|
| 115 |
+
"109": "truck",
|
| 116 |
+
"110": "trumpet",
|
| 117 |
+
"111": "umbrella",
|
| 118 |
+
"112": "vase",
|
| 119 |
+
"113": "violin",
|
| 120 |
+
"114": "watermelon",
|
| 121 |
+
"115": "whale",
|
| 122 |
+
"116": "windmill",
|
| 123 |
+
"117": "wine_glass",
|
| 124 |
+
"118": "yoga",
|
| 125 |
+
"119": "zebra",
|
| 126 |
+
"120": "zigzag"
|
| 127 |
+
},
|
| 128 |
+
"initializer_factor": 1.0,
|
| 129 |
+
"label2id": {
|
| 130 |
+
"The_Eiffel_Tower": 97,
|
| 131 |
+
"The_Great_Wall_of_China": 98,
|
| 132 |
+
"The_Mona_Lisa": 99,
|
| 133 |
+
"barn": 0,
|
| 134 |
+
"baseball_bat": 1,
|
| 135 |
+
"basket": 2,
|
| 136 |
+
"beach": 3,
|
| 137 |
+
"bear": 4,
|
| 138 |
+
"beard": 5,
|
| 139 |
+
"bee": 6,
|
| 140 |
+
"bird": 7,
|
| 141 |
+
"blueberry": 8,
|
| 142 |
+
"bowtie": 9,
|
| 143 |
+
"bracelet": 10,
|
| 144 |
+
"brain": 11,
|
| 145 |
+
"bread": 12,
|
| 146 |
+
"broccoli": 13,
|
| 147 |
+
"bus": 14,
|
| 148 |
+
"butterfly": 15,
|
| 149 |
+
"circle": 16,
|
| 150 |
+
"cloud": 17,
|
| 151 |
+
"cruise_ship": 18,
|
| 152 |
+
"dolphin": 19,
|
| 153 |
+
"dumbbell": 20,
|
| 154 |
+
"elephant": 21,
|
| 155 |
+
"eye": 22,
|
| 156 |
+
"eyeglasses": 23,
|
| 157 |
+
"feather": 24,
|
| 158 |
+
"fish": 25,
|
| 159 |
+
"flower": 26,
|
| 160 |
+
"foot": 27,
|
| 161 |
+
"frog": 28,
|
| 162 |
+
"giraffe": 29,
|
| 163 |
+
"goatee": 30,
|
| 164 |
+
"golf_club": 31,
|
| 165 |
+
"grapes": 32,
|
| 166 |
+
"grass": 33,
|
| 167 |
+
"guitar": 34,
|
| 168 |
+
"hamburger": 35,
|
| 169 |
+
"hand": 36,
|
| 170 |
+
"hat": 37,
|
| 171 |
+
"headphones": 38,
|
| 172 |
+
"helicopter": 39,
|
| 173 |
+
"hexagon": 40,
|
| 174 |
+
"hockey_stick": 41,
|
| 175 |
+
"horse": 42,
|
| 176 |
+
"hourglass": 43,
|
| 177 |
+
"house": 44,
|
| 178 |
+
"ice_cream": 45,
|
| 179 |
+
"jacket": 46,
|
| 180 |
+
"ladder": 47,
|
| 181 |
+
"leg": 48,
|
| 182 |
+
"lipstick": 49,
|
| 183 |
+
"megaphone": 50,
|
| 184 |
+
"monkey": 51,
|
| 185 |
+
"moon": 52,
|
| 186 |
+
"mushroom": 53,
|
| 187 |
+
"necklace": 54,
|
| 188 |
+
"owl": 55,
|
| 189 |
+
"panda": 56,
|
| 190 |
+
"pear": 57,
|
| 191 |
+
"peas": 58,
|
| 192 |
+
"penguin": 59,
|
| 193 |
+
"pig": 60,
|
| 194 |
+
"pillow": 61,
|
| 195 |
+
"pineapple": 62,
|
| 196 |
+
"pizza": 63,
|
| 197 |
+
"pool": 64,
|
| 198 |
+
"popsicle": 65,
|
| 199 |
+
"rabbit": 66,
|
| 200 |
+
"rhinoceros": 67,
|
| 201 |
+
"rifle": 68,
|
| 202 |
+
"river": 69,
|
| 203 |
+
"sailboat": 70,
|
| 204 |
+
"sandwich": 71,
|
| 205 |
+
"sea_turtle": 72,
|
| 206 |
+
"shark": 73,
|
| 207 |
+
"shoe": 74,
|
| 208 |
+
"skyscraper": 75,
|
| 209 |
+
"snorkel": 76,
|
| 210 |
+
"snowman": 77,
|
| 211 |
+
"soccer_ball": 78,
|
| 212 |
+
"speedboat": 79,
|
| 213 |
+
"spider": 80,
|
| 214 |
+
"spoon": 81,
|
| 215 |
+
"square": 82,
|
| 216 |
+
"squirrel": 83,
|
| 217 |
+
"stethoscope": 84,
|
| 218 |
+
"strawberry": 85,
|
| 219 |
+
"streetlight": 86,
|
| 220 |
+
"submarine": 87,
|
| 221 |
+
"suitcase": 88,
|
| 222 |
+
"sun": 89,
|
| 223 |
+
"sweater": 90,
|
| 224 |
+
"sword": 91,
|
| 225 |
+
"table": 92,
|
| 226 |
+
"teapot": 93,
|
| 227 |
+
"teddy-bear": 94,
|
| 228 |
+
"telephone": 95,
|
| 229 |
+
"tent": 96,
|
| 230 |
+
"tiger": 100,
|
| 231 |
+
"toaster": 101,
|
| 232 |
+
"tooth": 102,
|
| 233 |
+
"tornado": 103,
|
| 234 |
+
"tractor": 104,
|
| 235 |
+
"train": 105,
|
| 236 |
+
"tree": 106,
|
| 237 |
+
"triangle": 107,
|
| 238 |
+
"trombone": 108,
|
| 239 |
+
"truck": 109,
|
| 240 |
+
"trumpet": 110,
|
| 241 |
+
"umbrella": 111,
|
| 242 |
+
"vase": 112,
|
| 243 |
+
"violin": 113,
|
| 244 |
+
"watermelon": 114,
|
| 245 |
+
"whale": 115,
|
| 246 |
+
"windmill": 116,
|
| 247 |
+
"wine_glass": 117,
|
| 248 |
+
"yoga": 118,
|
| 249 |
+
"zebra": 119,
|
| 250 |
+
"zigzag": 120
|
| 251 |
+
},
|
| 252 |
+
"model_type": "siglip",
|
| 253 |
+
"problem_type": "single_label_classification",
|
| 254 |
+
"text_config": {
|
| 255 |
+
"attention_dropout": 0.0,
|
| 256 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 257 |
+
"hidden_size": 768,
|
| 258 |
+
"intermediate_size": 3072,
|
| 259 |
+
"layer_norm_eps": 1e-06,
|
| 260 |
+
"max_position_embeddings": 64,
|
| 261 |
+
"model_type": "siglip_text_model",
|
| 262 |
+
"num_attention_heads": 12,
|
| 263 |
+
"num_hidden_layers": 12,
|
| 264 |
+
"projection_size": 768,
|
| 265 |
+
"torch_dtype": "float32",
|
| 266 |
+
"vocab_size": 256000
|
| 267 |
+
},
|
| 268 |
+
"torch_dtype": "float32",
|
| 269 |
+
"transformers_version": "4.51.0.dev0",
|
| 270 |
+
"vision_config": {
|
| 271 |
+
"attention_dropout": 0.0,
|
| 272 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 273 |
+
"hidden_size": 768,
|
| 274 |
+
"image_size": 224,
|
| 275 |
+
"intermediate_size": 3072,
|
| 276 |
+
"layer_norm_eps": 1e-06,
|
| 277 |
+
"model_type": "siglip_vision_model",
|
| 278 |
+
"num_attention_heads": 12,
|
| 279 |
+
"num_channels": 3,
|
| 280 |
+
"num_hidden_layers": 12,
|
| 281 |
+
"patch_size": 16,
|
| 282 |
+
"torch_dtype": "float32"
|
| 283 |
+
}
|
| 284 |
+
}
|
checkpoint-3063/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a48511f8041ed23d8ef6b866fe8caa56e19c19323d43b36de4b39b7da0a53f3d
|
| 3 |
+
size 371934052
|
checkpoint-3063/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1fdbf49d7bfd2df4c5398d66c82285e8c56465af89eea03dc4359a18ac7c3612
|
| 3 |
+
size 687287802
|
checkpoint-3063/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-3063/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:211f72ee1ac51da6f7ccb2570cdb4e30234b4126bf39dfde63ec71758da621a5
|
| 3 |
+
size 14244
|
checkpoint-3063/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7548d314b3526255122b39d5cd6162a5b057112b63c2fc571557a7b4a6d6d36e
|
| 3 |
+
size 1064
|
checkpoint-3063/trainer_state.json
ADDED
|
@@ -0,0 +1,106 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 3063,
|
| 3 |
+
"best_metric": 0.9870542287826538,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-3063",
|
| 5 |
+
"epoch": 3.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 3063,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.48971596474045054,
|
| 14 |
+
"grad_norm": 59.27558135986328,
|
| 15 |
+
"learning_rate": 2.665344571145265e-06,
|
| 16 |
+
"loss": 4.268,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.9794319294809011,
|
| 21 |
+
"grad_norm": 61.66608810424805,
|
| 22 |
+
"learning_rate": 2.2935052057511156e-06,
|
| 23 |
+
"loss": 2.4217,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.666574839302112,
|
| 29 |
+
"eval_loss": 1.7632426023483276,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 476.9192,
|
| 32 |
+
"eval_samples_per_second": 68.502,
|
| 33 |
+
"eval_steps_per_second": 8.563,
|
| 34 |
+
"step": 1021
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 1.4691478942213516,
|
| 38 |
+
"grad_norm": 71.97106170654297,
|
| 39 |
+
"learning_rate": 1.921665840356966e-06,
|
| 40 |
+
"loss": 1.709,
|
| 41 |
+
"step": 1500
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 1.9588638589618022,
|
| 45 |
+
"grad_norm": 63.00515365600586,
|
| 46 |
+
"learning_rate": 1.5498264749628163e-06,
|
| 47 |
+
"loss": 1.4697,
|
| 48 |
+
"step": 2000
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 2.0,
|
| 52 |
+
"eval_accuracy": 0.7615243342516069,
|
| 53 |
+
"eval_loss": 1.171113133430481,
|
| 54 |
+
"eval_model_preparation_time": 0.0024,
|
| 55 |
+
"eval_runtime": 475.7496,
|
| 56 |
+
"eval_samples_per_second": 68.671,
|
| 57 |
+
"eval_steps_per_second": 8.584,
|
| 58 |
+
"step": 2042
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 2.4485798237022527,
|
| 62 |
+
"grad_norm": 51.789241790771484,
|
| 63 |
+
"learning_rate": 1.1779871095686663e-06,
|
| 64 |
+
"loss": 1.2603,
|
| 65 |
+
"step": 2500
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 2.9382957884427032,
|
| 69 |
+
"grad_norm": 42.61498260498047,
|
| 70 |
+
"learning_rate": 8.061477441745166e-07,
|
| 71 |
+
"loss": 1.2054,
|
| 72 |
+
"step": 3000
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 3.0,
|
| 76 |
+
"eval_accuracy": 0.7891031527395164,
|
| 77 |
+
"eval_loss": 0.9870542287826538,
|
| 78 |
+
"eval_model_preparation_time": 0.0024,
|
| 79 |
+
"eval_runtime": 475.3188,
|
| 80 |
+
"eval_samples_per_second": 68.733,
|
| 81 |
+
"eval_steps_per_second": 8.592,
|
| 82 |
+
"step": 3063
|
| 83 |
+
}
|
| 84 |
+
],
|
| 85 |
+
"logging_steps": 500,
|
| 86 |
+
"max_steps": 4084,
|
| 87 |
+
"num_input_tokens_seen": 0,
|
| 88 |
+
"num_train_epochs": 4,
|
| 89 |
+
"save_steps": 500,
|
| 90 |
+
"stateful_callbacks": {
|
| 91 |
+
"TrainerControl": {
|
| 92 |
+
"args": {
|
| 93 |
+
"should_epoch_stop": false,
|
| 94 |
+
"should_evaluate": false,
|
| 95 |
+
"should_log": false,
|
| 96 |
+
"should_save": true,
|
| 97 |
+
"should_training_stop": false
|
| 98 |
+
},
|
| 99 |
+
"attributes": {}
|
| 100 |
+
}
|
| 101 |
+
},
|
| 102 |
+
"total_flos": 8.216976653015962e+18,
|
| 103 |
+
"train_batch_size": 32,
|
| 104 |
+
"trial_name": null,
|
| 105 |
+
"trial_params": null
|
| 106 |
+
}
|
checkpoint-3063/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eff7895602b654dfe6fc27f9bfd70976ed586448b1d172d1abb6cf2d3bccf43
|
| 3 |
+
size 5304
|
checkpoint-4084/config.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "barn",
|
| 7 |
+
"1": "baseball_bat",
|
| 8 |
+
"2": "basket",
|
| 9 |
+
"3": "beach",
|
| 10 |
+
"4": "bear",
|
| 11 |
+
"5": "beard",
|
| 12 |
+
"6": "bee",
|
| 13 |
+
"7": "bird",
|
| 14 |
+
"8": "blueberry",
|
| 15 |
+
"9": "bowtie",
|
| 16 |
+
"10": "bracelet",
|
| 17 |
+
"11": "brain",
|
| 18 |
+
"12": "bread",
|
| 19 |
+
"13": "broccoli",
|
| 20 |
+
"14": "bus",
|
| 21 |
+
"15": "butterfly",
|
| 22 |
+
"16": "circle",
|
| 23 |
+
"17": "cloud",
|
| 24 |
+
"18": "cruise_ship",
|
| 25 |
+
"19": "dolphin",
|
| 26 |
+
"20": "dumbbell",
|
| 27 |
+
"21": "elephant",
|
| 28 |
+
"22": "eye",
|
| 29 |
+
"23": "eyeglasses",
|
| 30 |
+
"24": "feather",
|
| 31 |
+
"25": "fish",
|
| 32 |
+
"26": "flower",
|
| 33 |
+
"27": "foot",
|
| 34 |
+
"28": "frog",
|
| 35 |
+
"29": "giraffe",
|
| 36 |
+
"30": "goatee",
|
| 37 |
+
"31": "golf_club",
|
| 38 |
+
"32": "grapes",
|
| 39 |
+
"33": "grass",
|
| 40 |
+
"34": "guitar",
|
| 41 |
+
"35": "hamburger",
|
| 42 |
+
"36": "hand",
|
| 43 |
+
"37": "hat",
|
| 44 |
+
"38": "headphones",
|
| 45 |
+
"39": "helicopter",
|
| 46 |
+
"40": "hexagon",
|
| 47 |
+
"41": "hockey_stick",
|
| 48 |
+
"42": "horse",
|
| 49 |
+
"43": "hourglass",
|
| 50 |
+
"44": "house",
|
| 51 |
+
"45": "ice_cream",
|
| 52 |
+
"46": "jacket",
|
| 53 |
+
"47": "ladder",
|
| 54 |
+
"48": "leg",
|
| 55 |
+
"49": "lipstick",
|
| 56 |
+
"50": "megaphone",
|
| 57 |
+
"51": "monkey",
|
| 58 |
+
"52": "moon",
|
| 59 |
+
"53": "mushroom",
|
| 60 |
+
"54": "necklace",
|
| 61 |
+
"55": "owl",
|
| 62 |
+
"56": "panda",
|
| 63 |
+
"57": "pear",
|
| 64 |
+
"58": "peas",
|
| 65 |
+
"59": "penguin",
|
| 66 |
+
"60": "pig",
|
| 67 |
+
"61": "pillow",
|
| 68 |
+
"62": "pineapple",
|
| 69 |
+
"63": "pizza",
|
| 70 |
+
"64": "pool",
|
| 71 |
+
"65": "popsicle",
|
| 72 |
+
"66": "rabbit",
|
| 73 |
+
"67": "rhinoceros",
|
| 74 |
+
"68": "rifle",
|
| 75 |
+
"69": "river",
|
| 76 |
+
"70": "sailboat",
|
| 77 |
+
"71": "sandwich",
|
| 78 |
+
"72": "sea_turtle",
|
| 79 |
+
"73": "shark",
|
| 80 |
+
"74": "shoe",
|
| 81 |
+
"75": "skyscraper",
|
| 82 |
+
"76": "snorkel",
|
| 83 |
+
"77": "snowman",
|
| 84 |
+
"78": "soccer_ball",
|
| 85 |
+
"79": "speedboat",
|
| 86 |
+
"80": "spider",
|
| 87 |
+
"81": "spoon",
|
| 88 |
+
"82": "square",
|
| 89 |
+
"83": "squirrel",
|
| 90 |
+
"84": "stethoscope",
|
| 91 |
+
"85": "strawberry",
|
| 92 |
+
"86": "streetlight",
|
| 93 |
+
"87": "submarine",
|
| 94 |
+
"88": "suitcase",
|
| 95 |
+
"89": "sun",
|
| 96 |
+
"90": "sweater",
|
| 97 |
+
"91": "sword",
|
| 98 |
+
"92": "table",
|
| 99 |
+
"93": "teapot",
|
| 100 |
+
"94": "teddy-bear",
|
| 101 |
+
"95": "telephone",
|
| 102 |
+
"96": "tent",
|
| 103 |
+
"97": "The_Eiffel_Tower",
|
| 104 |
+
"98": "The_Great_Wall_of_China",
|
| 105 |
+
"99": "The_Mona_Lisa",
|
| 106 |
+
"100": "tiger",
|
| 107 |
+
"101": "toaster",
|
| 108 |
+
"102": "tooth",
|
| 109 |
+
"103": "tornado",
|
| 110 |
+
"104": "tractor",
|
| 111 |
+
"105": "train",
|
| 112 |
+
"106": "tree",
|
| 113 |
+
"107": "triangle",
|
| 114 |
+
"108": "trombone",
|
| 115 |
+
"109": "truck",
|
| 116 |
+
"110": "trumpet",
|
| 117 |
+
"111": "umbrella",
|
| 118 |
+
"112": "vase",
|
| 119 |
+
"113": "violin",
|
| 120 |
+
"114": "watermelon",
|
| 121 |
+
"115": "whale",
|
| 122 |
+
"116": "windmill",
|
| 123 |
+
"117": "wine_glass",
|
| 124 |
+
"118": "yoga",
|
| 125 |
+
"119": "zebra",
|
| 126 |
+
"120": "zigzag"
|
| 127 |
+
},
|
| 128 |
+
"initializer_factor": 1.0,
|
| 129 |
+
"label2id": {
|
| 130 |
+
"The_Eiffel_Tower": 97,
|
| 131 |
+
"The_Great_Wall_of_China": 98,
|
| 132 |
+
"The_Mona_Lisa": 99,
|
| 133 |
+
"barn": 0,
|
| 134 |
+
"baseball_bat": 1,
|
| 135 |
+
"basket": 2,
|
| 136 |
+
"beach": 3,
|
| 137 |
+
"bear": 4,
|
| 138 |
+
"beard": 5,
|
| 139 |
+
"bee": 6,
|
| 140 |
+
"bird": 7,
|
| 141 |
+
"blueberry": 8,
|
| 142 |
+
"bowtie": 9,
|
| 143 |
+
"bracelet": 10,
|
| 144 |
+
"brain": 11,
|
| 145 |
+
"bread": 12,
|
| 146 |
+
"broccoli": 13,
|
| 147 |
+
"bus": 14,
|
| 148 |
+
"butterfly": 15,
|
| 149 |
+
"circle": 16,
|
| 150 |
+
"cloud": 17,
|
| 151 |
+
"cruise_ship": 18,
|
| 152 |
+
"dolphin": 19,
|
| 153 |
+
"dumbbell": 20,
|
| 154 |
+
"elephant": 21,
|
| 155 |
+
"eye": 22,
|
| 156 |
+
"eyeglasses": 23,
|
| 157 |
+
"feather": 24,
|
| 158 |
+
"fish": 25,
|
| 159 |
+
"flower": 26,
|
| 160 |
+
"foot": 27,
|
| 161 |
+
"frog": 28,
|
| 162 |
+
"giraffe": 29,
|
| 163 |
+
"goatee": 30,
|
| 164 |
+
"golf_club": 31,
|
| 165 |
+
"grapes": 32,
|
| 166 |
+
"grass": 33,
|
| 167 |
+
"guitar": 34,
|
| 168 |
+
"hamburger": 35,
|
| 169 |
+
"hand": 36,
|
| 170 |
+
"hat": 37,
|
| 171 |
+
"headphones": 38,
|
| 172 |
+
"helicopter": 39,
|
| 173 |
+
"hexagon": 40,
|
| 174 |
+
"hockey_stick": 41,
|
| 175 |
+
"horse": 42,
|
| 176 |
+
"hourglass": 43,
|
| 177 |
+
"house": 44,
|
| 178 |
+
"ice_cream": 45,
|
| 179 |
+
"jacket": 46,
|
| 180 |
+
"ladder": 47,
|
| 181 |
+
"leg": 48,
|
| 182 |
+
"lipstick": 49,
|
| 183 |
+
"megaphone": 50,
|
| 184 |
+
"monkey": 51,
|
| 185 |
+
"moon": 52,
|
| 186 |
+
"mushroom": 53,
|
| 187 |
+
"necklace": 54,
|
| 188 |
+
"owl": 55,
|
| 189 |
+
"panda": 56,
|
| 190 |
+
"pear": 57,
|
| 191 |
+
"peas": 58,
|
| 192 |
+
"penguin": 59,
|
| 193 |
+
"pig": 60,
|
| 194 |
+
"pillow": 61,
|
| 195 |
+
"pineapple": 62,
|
| 196 |
+
"pizza": 63,
|
| 197 |
+
"pool": 64,
|
| 198 |
+
"popsicle": 65,
|
| 199 |
+
"rabbit": 66,
|
| 200 |
+
"rhinoceros": 67,
|
| 201 |
+
"rifle": 68,
|
| 202 |
+
"river": 69,
|
| 203 |
+
"sailboat": 70,
|
| 204 |
+
"sandwich": 71,
|
| 205 |
+
"sea_turtle": 72,
|
| 206 |
+
"shark": 73,
|
| 207 |
+
"shoe": 74,
|
| 208 |
+
"skyscraper": 75,
|
| 209 |
+
"snorkel": 76,
|
| 210 |
+
"snowman": 77,
|
| 211 |
+
"soccer_ball": 78,
|
| 212 |
+
"speedboat": 79,
|
| 213 |
+
"spider": 80,
|
| 214 |
+
"spoon": 81,
|
| 215 |
+
"square": 82,
|
| 216 |
+
"squirrel": 83,
|
| 217 |
+
"stethoscope": 84,
|
| 218 |
+
"strawberry": 85,
|
| 219 |
+
"streetlight": 86,
|
| 220 |
+
"submarine": 87,
|
| 221 |
+
"suitcase": 88,
|
| 222 |
+
"sun": 89,
|
| 223 |
+
"sweater": 90,
|
| 224 |
+
"sword": 91,
|
| 225 |
+
"table": 92,
|
| 226 |
+
"teapot": 93,
|
| 227 |
+
"teddy-bear": 94,
|
| 228 |
+
"telephone": 95,
|
| 229 |
+
"tent": 96,
|
| 230 |
+
"tiger": 100,
|
| 231 |
+
"toaster": 101,
|
| 232 |
+
"tooth": 102,
|
| 233 |
+
"tornado": 103,
|
| 234 |
+
"tractor": 104,
|
| 235 |
+
"train": 105,
|
| 236 |
+
"tree": 106,
|
| 237 |
+
"triangle": 107,
|
| 238 |
+
"trombone": 108,
|
| 239 |
+
"truck": 109,
|
| 240 |
+
"trumpet": 110,
|
| 241 |
+
"umbrella": 111,
|
| 242 |
+
"vase": 112,
|
| 243 |
+
"violin": 113,
|
| 244 |
+
"watermelon": 114,
|
| 245 |
+
"whale": 115,
|
| 246 |
+
"windmill": 116,
|
| 247 |
+
"wine_glass": 117,
|
| 248 |
+
"yoga": 118,
|
| 249 |
+
"zebra": 119,
|
| 250 |
+
"zigzag": 120
|
| 251 |
+
},
|
| 252 |
+
"model_type": "siglip",
|
| 253 |
+
"problem_type": "single_label_classification",
|
| 254 |
+
"text_config": {
|
| 255 |
+
"attention_dropout": 0.0,
|
| 256 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 257 |
+
"hidden_size": 768,
|
| 258 |
+
"intermediate_size": 3072,
|
| 259 |
+
"layer_norm_eps": 1e-06,
|
| 260 |
+
"max_position_embeddings": 64,
|
| 261 |
+
"model_type": "siglip_text_model",
|
| 262 |
+
"num_attention_heads": 12,
|
| 263 |
+
"num_hidden_layers": 12,
|
| 264 |
+
"projection_size": 768,
|
| 265 |
+
"torch_dtype": "float32",
|
| 266 |
+
"vocab_size": 256000
|
| 267 |
+
},
|
| 268 |
+
"torch_dtype": "float32",
|
| 269 |
+
"transformers_version": "4.51.0.dev0",
|
| 270 |
+
"vision_config": {
|
| 271 |
+
"attention_dropout": 0.0,
|
| 272 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 273 |
+
"hidden_size": 768,
|
| 274 |
+
"image_size": 224,
|
| 275 |
+
"intermediate_size": 3072,
|
| 276 |
+
"layer_norm_eps": 1e-06,
|
| 277 |
+
"model_type": "siglip_vision_model",
|
| 278 |
+
"num_attention_heads": 12,
|
| 279 |
+
"num_channels": 3,
|
| 280 |
+
"num_hidden_layers": 12,
|
| 281 |
+
"patch_size": 16,
|
| 282 |
+
"torch_dtype": "float32"
|
| 283 |
+
}
|
| 284 |
+
}
|
checkpoint-4084/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8edc9977470158b413298172b9b0fc94d7aa54298252c799de641d44acb763c6
|
| 3 |
+
size 371934052
|
checkpoint-4084/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:21cadc3833dfffbc9d4ef27ab50c1322e366e8b755e1ebf54fa2a89793353376
|
| 3 |
+
size 687287802
|
checkpoint-4084/preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
checkpoint-4084/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eedc7612b1b138c3d72a61e0ada7b545950da37e80751dabafdbeb3c9f2121ba
|
| 3 |
+
size 14244
|
checkpoint-4084/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dcc7a84b21fc1a6bc91d3adfc0a2618738711a256c159abf62175373c1770d0e
|
| 3 |
+
size 1064
|
checkpoint-4084/trainer_state.json
ADDED
|
@@ -0,0 +1,130 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 4084,
|
| 3 |
+
"best_metric": 0.9286365509033203,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-4084",
|
| 5 |
+
"epoch": 4.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 4084,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.48971596474045054,
|
| 14 |
+
"grad_norm": 59.27558135986328,
|
| 15 |
+
"learning_rate": 2.665344571145265e-06,
|
| 16 |
+
"loss": 4.268,
|
| 17 |
+
"step": 500
|
| 18 |
+
},
|
| 19 |
+
{
|
| 20 |
+
"epoch": 0.9794319294809011,
|
| 21 |
+
"grad_norm": 61.66608810424805,
|
| 22 |
+
"learning_rate": 2.2935052057511156e-06,
|
| 23 |
+
"loss": 2.4217,
|
| 24 |
+
"step": 1000
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 1.0,
|
| 28 |
+
"eval_accuracy": 0.666574839302112,
|
| 29 |
+
"eval_loss": 1.7632426023483276,
|
| 30 |
+
"eval_model_preparation_time": 0.0024,
|
| 31 |
+
"eval_runtime": 476.9192,
|
| 32 |
+
"eval_samples_per_second": 68.502,
|
| 33 |
+
"eval_steps_per_second": 8.563,
|
| 34 |
+
"step": 1021
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 1.4691478942213516,
|
| 38 |
+
"grad_norm": 71.97106170654297,
|
| 39 |
+
"learning_rate": 1.921665840356966e-06,
|
| 40 |
+
"loss": 1.709,
|
| 41 |
+
"step": 1500
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 1.9588638589618022,
|
| 45 |
+
"grad_norm": 63.00515365600586,
|
| 46 |
+
"learning_rate": 1.5498264749628163e-06,
|
| 47 |
+
"loss": 1.4697,
|
| 48 |
+
"step": 2000
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 2.0,
|
| 52 |
+
"eval_accuracy": 0.7615243342516069,
|
| 53 |
+
"eval_loss": 1.171113133430481,
|
| 54 |
+
"eval_model_preparation_time": 0.0024,
|
| 55 |
+
"eval_runtime": 475.7496,
|
| 56 |
+
"eval_samples_per_second": 68.671,
|
| 57 |
+
"eval_steps_per_second": 8.584,
|
| 58 |
+
"step": 2042
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 2.4485798237022527,
|
| 62 |
+
"grad_norm": 51.789241790771484,
|
| 63 |
+
"learning_rate": 1.1779871095686663e-06,
|
| 64 |
+
"loss": 1.2603,
|
| 65 |
+
"step": 2500
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 2.9382957884427032,
|
| 69 |
+
"grad_norm": 42.61498260498047,
|
| 70 |
+
"learning_rate": 8.061477441745166e-07,
|
| 71 |
+
"loss": 1.2054,
|
| 72 |
+
"step": 3000
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 3.0,
|
| 76 |
+
"eval_accuracy": 0.7891031527395164,
|
| 77 |
+
"eval_loss": 0.9870542287826538,
|
| 78 |
+
"eval_model_preparation_time": 0.0024,
|
| 79 |
+
"eval_runtime": 475.3188,
|
| 80 |
+
"eval_samples_per_second": 68.733,
|
| 81 |
+
"eval_steps_per_second": 8.592,
|
| 82 |
+
"step": 3063
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"epoch": 3.4280117531831538,
|
| 86 |
+
"grad_norm": 50.29034423828125,
|
| 87 |
+
"learning_rate": 4.343083787803669e-07,
|
| 88 |
+
"loss": 1.1072,
|
| 89 |
+
"step": 3500
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 3.9177277179236043,
|
| 93 |
+
"grad_norm": 67.87676239013672,
|
| 94 |
+
"learning_rate": 6.246901338621716e-08,
|
| 95 |
+
"loss": 1.0737,
|
| 96 |
+
"step": 4000
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"epoch": 4.0,
|
| 100 |
+
"eval_accuracy": 0.7995102540557086,
|
| 101 |
+
"eval_loss": 0.9286365509033203,
|
| 102 |
+
"eval_model_preparation_time": 0.0024,
|
| 103 |
+
"eval_runtime": 478.0209,
|
| 104 |
+
"eval_samples_per_second": 68.344,
|
| 105 |
+
"eval_steps_per_second": 8.544,
|
| 106 |
+
"step": 4084
|
| 107 |
+
}
|
| 108 |
+
],
|
| 109 |
+
"logging_steps": 500,
|
| 110 |
+
"max_steps": 4084,
|
| 111 |
+
"num_input_tokens_seen": 0,
|
| 112 |
+
"num_train_epochs": 4,
|
| 113 |
+
"save_steps": 500,
|
| 114 |
+
"stateful_callbacks": {
|
| 115 |
+
"TrainerControl": {
|
| 116 |
+
"args": {
|
| 117 |
+
"should_epoch_stop": false,
|
| 118 |
+
"should_evaluate": false,
|
| 119 |
+
"should_log": false,
|
| 120 |
+
"should_save": true,
|
| 121 |
+
"should_training_stop": true
|
| 122 |
+
},
|
| 123 |
+
"attributes": {}
|
| 124 |
+
}
|
| 125 |
+
},
|
| 126 |
+
"total_flos": 1.0955968870687949e+19,
|
| 127 |
+
"train_batch_size": 32,
|
| 128 |
+
"trial_name": null,
|
| 129 |
+
"trial_params": null
|
| 130 |
+
}
|
checkpoint-4084/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eff7895602b654dfe6fc27f9bfd70976ed586448b1d172d1abb6cf2d3bccf43
|
| 3 |
+
size 5304
|
config.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "barn",
|
| 7 |
+
"1": "baseball_bat",
|
| 8 |
+
"2": "basket",
|
| 9 |
+
"3": "beach",
|
| 10 |
+
"4": "bear",
|
| 11 |
+
"5": "beard",
|
| 12 |
+
"6": "bee",
|
| 13 |
+
"7": "bird",
|
| 14 |
+
"8": "blueberry",
|
| 15 |
+
"9": "bowtie",
|
| 16 |
+
"10": "bracelet",
|
| 17 |
+
"11": "brain",
|
| 18 |
+
"12": "bread",
|
| 19 |
+
"13": "broccoli",
|
| 20 |
+
"14": "bus",
|
| 21 |
+
"15": "butterfly",
|
| 22 |
+
"16": "circle",
|
| 23 |
+
"17": "cloud",
|
| 24 |
+
"18": "cruise_ship",
|
| 25 |
+
"19": "dolphin",
|
| 26 |
+
"20": "dumbbell",
|
| 27 |
+
"21": "elephant",
|
| 28 |
+
"22": "eye",
|
| 29 |
+
"23": "eyeglasses",
|
| 30 |
+
"24": "feather",
|
| 31 |
+
"25": "fish",
|
| 32 |
+
"26": "flower",
|
| 33 |
+
"27": "foot",
|
| 34 |
+
"28": "frog",
|
| 35 |
+
"29": "giraffe",
|
| 36 |
+
"30": "goatee",
|
| 37 |
+
"31": "golf_club",
|
| 38 |
+
"32": "grapes",
|
| 39 |
+
"33": "grass",
|
| 40 |
+
"34": "guitar",
|
| 41 |
+
"35": "hamburger",
|
| 42 |
+
"36": "hand",
|
| 43 |
+
"37": "hat",
|
| 44 |
+
"38": "headphones",
|
| 45 |
+
"39": "helicopter",
|
| 46 |
+
"40": "hexagon",
|
| 47 |
+
"41": "hockey_stick",
|
| 48 |
+
"42": "horse",
|
| 49 |
+
"43": "hourglass",
|
| 50 |
+
"44": "house",
|
| 51 |
+
"45": "ice_cream",
|
| 52 |
+
"46": "jacket",
|
| 53 |
+
"47": "ladder",
|
| 54 |
+
"48": "leg",
|
| 55 |
+
"49": "lipstick",
|
| 56 |
+
"50": "megaphone",
|
| 57 |
+
"51": "monkey",
|
| 58 |
+
"52": "moon",
|
| 59 |
+
"53": "mushroom",
|
| 60 |
+
"54": "necklace",
|
| 61 |
+
"55": "owl",
|
| 62 |
+
"56": "panda",
|
| 63 |
+
"57": "pear",
|
| 64 |
+
"58": "peas",
|
| 65 |
+
"59": "penguin",
|
| 66 |
+
"60": "pig",
|
| 67 |
+
"61": "pillow",
|
| 68 |
+
"62": "pineapple",
|
| 69 |
+
"63": "pizza",
|
| 70 |
+
"64": "pool",
|
| 71 |
+
"65": "popsicle",
|
| 72 |
+
"66": "rabbit",
|
| 73 |
+
"67": "rhinoceros",
|
| 74 |
+
"68": "rifle",
|
| 75 |
+
"69": "river",
|
| 76 |
+
"70": "sailboat",
|
| 77 |
+
"71": "sandwich",
|
| 78 |
+
"72": "sea_turtle",
|
| 79 |
+
"73": "shark",
|
| 80 |
+
"74": "shoe",
|
| 81 |
+
"75": "skyscraper",
|
| 82 |
+
"76": "snorkel",
|
| 83 |
+
"77": "snowman",
|
| 84 |
+
"78": "soccer_ball",
|
| 85 |
+
"79": "speedboat",
|
| 86 |
+
"80": "spider",
|
| 87 |
+
"81": "spoon",
|
| 88 |
+
"82": "square",
|
| 89 |
+
"83": "squirrel",
|
| 90 |
+
"84": "stethoscope",
|
| 91 |
+
"85": "strawberry",
|
| 92 |
+
"86": "streetlight",
|
| 93 |
+
"87": "submarine",
|
| 94 |
+
"88": "suitcase",
|
| 95 |
+
"89": "sun",
|
| 96 |
+
"90": "sweater",
|
| 97 |
+
"91": "sword",
|
| 98 |
+
"92": "table",
|
| 99 |
+
"93": "teapot",
|
| 100 |
+
"94": "teddy-bear",
|
| 101 |
+
"95": "telephone",
|
| 102 |
+
"96": "tent",
|
| 103 |
+
"97": "The_Eiffel_Tower",
|
| 104 |
+
"98": "The_Great_Wall_of_China",
|
| 105 |
+
"99": "The_Mona_Lisa",
|
| 106 |
+
"100": "tiger",
|
| 107 |
+
"101": "toaster",
|
| 108 |
+
"102": "tooth",
|
| 109 |
+
"103": "tornado",
|
| 110 |
+
"104": "tractor",
|
| 111 |
+
"105": "train",
|
| 112 |
+
"106": "tree",
|
| 113 |
+
"107": "triangle",
|
| 114 |
+
"108": "trombone",
|
| 115 |
+
"109": "truck",
|
| 116 |
+
"110": "trumpet",
|
| 117 |
+
"111": "umbrella",
|
| 118 |
+
"112": "vase",
|
| 119 |
+
"113": "violin",
|
| 120 |
+
"114": "watermelon",
|
| 121 |
+
"115": "whale",
|
| 122 |
+
"116": "windmill",
|
| 123 |
+
"117": "wine_glass",
|
| 124 |
+
"118": "yoga",
|
| 125 |
+
"119": "zebra",
|
| 126 |
+
"120": "zigzag"
|
| 127 |
+
},
|
| 128 |
+
"initializer_factor": 1.0,
|
| 129 |
+
"label2id": {
|
| 130 |
+
"The_Eiffel_Tower": 97,
|
| 131 |
+
"The_Great_Wall_of_China": 98,
|
| 132 |
+
"The_Mona_Lisa": 99,
|
| 133 |
+
"barn": 0,
|
| 134 |
+
"baseball_bat": 1,
|
| 135 |
+
"basket": 2,
|
| 136 |
+
"beach": 3,
|
| 137 |
+
"bear": 4,
|
| 138 |
+
"beard": 5,
|
| 139 |
+
"bee": 6,
|
| 140 |
+
"bird": 7,
|
| 141 |
+
"blueberry": 8,
|
| 142 |
+
"bowtie": 9,
|
| 143 |
+
"bracelet": 10,
|
| 144 |
+
"brain": 11,
|
| 145 |
+
"bread": 12,
|
| 146 |
+
"broccoli": 13,
|
| 147 |
+
"bus": 14,
|
| 148 |
+
"butterfly": 15,
|
| 149 |
+
"circle": 16,
|
| 150 |
+
"cloud": 17,
|
| 151 |
+
"cruise_ship": 18,
|
| 152 |
+
"dolphin": 19,
|
| 153 |
+
"dumbbell": 20,
|
| 154 |
+
"elephant": 21,
|
| 155 |
+
"eye": 22,
|
| 156 |
+
"eyeglasses": 23,
|
| 157 |
+
"feather": 24,
|
| 158 |
+
"fish": 25,
|
| 159 |
+
"flower": 26,
|
| 160 |
+
"foot": 27,
|
| 161 |
+
"frog": 28,
|
| 162 |
+
"giraffe": 29,
|
| 163 |
+
"goatee": 30,
|
| 164 |
+
"golf_club": 31,
|
| 165 |
+
"grapes": 32,
|
| 166 |
+
"grass": 33,
|
| 167 |
+
"guitar": 34,
|
| 168 |
+
"hamburger": 35,
|
| 169 |
+
"hand": 36,
|
| 170 |
+
"hat": 37,
|
| 171 |
+
"headphones": 38,
|
| 172 |
+
"helicopter": 39,
|
| 173 |
+
"hexagon": 40,
|
| 174 |
+
"hockey_stick": 41,
|
| 175 |
+
"horse": 42,
|
| 176 |
+
"hourglass": 43,
|
| 177 |
+
"house": 44,
|
| 178 |
+
"ice_cream": 45,
|
| 179 |
+
"jacket": 46,
|
| 180 |
+
"ladder": 47,
|
| 181 |
+
"leg": 48,
|
| 182 |
+
"lipstick": 49,
|
| 183 |
+
"megaphone": 50,
|
| 184 |
+
"monkey": 51,
|
| 185 |
+
"moon": 52,
|
| 186 |
+
"mushroom": 53,
|
| 187 |
+
"necklace": 54,
|
| 188 |
+
"owl": 55,
|
| 189 |
+
"panda": 56,
|
| 190 |
+
"pear": 57,
|
| 191 |
+
"peas": 58,
|
| 192 |
+
"penguin": 59,
|
| 193 |
+
"pig": 60,
|
| 194 |
+
"pillow": 61,
|
| 195 |
+
"pineapple": 62,
|
| 196 |
+
"pizza": 63,
|
| 197 |
+
"pool": 64,
|
| 198 |
+
"popsicle": 65,
|
| 199 |
+
"rabbit": 66,
|
| 200 |
+
"rhinoceros": 67,
|
| 201 |
+
"rifle": 68,
|
| 202 |
+
"river": 69,
|
| 203 |
+
"sailboat": 70,
|
| 204 |
+
"sandwich": 71,
|
| 205 |
+
"sea_turtle": 72,
|
| 206 |
+
"shark": 73,
|
| 207 |
+
"shoe": 74,
|
| 208 |
+
"skyscraper": 75,
|
| 209 |
+
"snorkel": 76,
|
| 210 |
+
"snowman": 77,
|
| 211 |
+
"soccer_ball": 78,
|
| 212 |
+
"speedboat": 79,
|
| 213 |
+
"spider": 80,
|
| 214 |
+
"spoon": 81,
|
| 215 |
+
"square": 82,
|
| 216 |
+
"squirrel": 83,
|
| 217 |
+
"stethoscope": 84,
|
| 218 |
+
"strawberry": 85,
|
| 219 |
+
"streetlight": 86,
|
| 220 |
+
"submarine": 87,
|
| 221 |
+
"suitcase": 88,
|
| 222 |
+
"sun": 89,
|
| 223 |
+
"sweater": 90,
|
| 224 |
+
"sword": 91,
|
| 225 |
+
"table": 92,
|
| 226 |
+
"teapot": 93,
|
| 227 |
+
"teddy-bear": 94,
|
| 228 |
+
"telephone": 95,
|
| 229 |
+
"tent": 96,
|
| 230 |
+
"tiger": 100,
|
| 231 |
+
"toaster": 101,
|
| 232 |
+
"tooth": 102,
|
| 233 |
+
"tornado": 103,
|
| 234 |
+
"tractor": 104,
|
| 235 |
+
"train": 105,
|
| 236 |
+
"tree": 106,
|
| 237 |
+
"triangle": 107,
|
| 238 |
+
"trombone": 108,
|
| 239 |
+
"truck": 109,
|
| 240 |
+
"trumpet": 110,
|
| 241 |
+
"umbrella": 111,
|
| 242 |
+
"vase": 112,
|
| 243 |
+
"violin": 113,
|
| 244 |
+
"watermelon": 114,
|
| 245 |
+
"whale": 115,
|
| 246 |
+
"windmill": 116,
|
| 247 |
+
"wine_glass": 117,
|
| 248 |
+
"yoga": 118,
|
| 249 |
+
"zebra": 119,
|
| 250 |
+
"zigzag": 120
|
| 251 |
+
},
|
| 252 |
+
"model_type": "siglip",
|
| 253 |
+
"problem_type": "single_label_classification",
|
| 254 |
+
"text_config": {
|
| 255 |
+
"attention_dropout": 0.0,
|
| 256 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 257 |
+
"hidden_size": 768,
|
| 258 |
+
"intermediate_size": 3072,
|
| 259 |
+
"layer_norm_eps": 1e-06,
|
| 260 |
+
"max_position_embeddings": 64,
|
| 261 |
+
"model_type": "siglip_text_model",
|
| 262 |
+
"num_attention_heads": 12,
|
| 263 |
+
"num_hidden_layers": 12,
|
| 264 |
+
"projection_size": 768,
|
| 265 |
+
"torch_dtype": "float32",
|
| 266 |
+
"vocab_size": 256000
|
| 267 |
+
},
|
| 268 |
+
"torch_dtype": "float32",
|
| 269 |
+
"transformers_version": "4.51.0.dev0",
|
| 270 |
+
"vision_config": {
|
| 271 |
+
"attention_dropout": 0.0,
|
| 272 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 273 |
+
"hidden_size": 768,
|
| 274 |
+
"image_size": 224,
|
| 275 |
+
"intermediate_size": 3072,
|
| 276 |
+
"layer_norm_eps": 1e-06,
|
| 277 |
+
"model_type": "siglip_vision_model",
|
| 278 |
+
"num_attention_heads": 12,
|
| 279 |
+
"num_channels": 3,
|
| 280 |
+
"num_hidden_layers": 12,
|
| 281 |
+
"patch_size": 16,
|
| 282 |
+
"torch_dtype": "float32"
|
| 283 |
+
}
|
| 284 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8edc9977470158b413298172b9b0fc94d7aa54298252c799de641d44acb763c6
|
| 3 |
+
size 371934052
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_convert_rgb": null,
|
| 3 |
+
"do_normalize": true,
|
| 4 |
+
"do_rescale": true,
|
| 5 |
+
"do_resize": true,
|
| 6 |
+
"image_mean": [
|
| 7 |
+
0.5,
|
| 8 |
+
0.5,
|
| 9 |
+
0.5
|
| 10 |
+
],
|
| 11 |
+
"image_processor_type": "SiglipImageProcessor",
|
| 12 |
+
"image_std": [
|
| 13 |
+
0.5,
|
| 14 |
+
0.5,
|
| 15 |
+
0.5
|
| 16 |
+
],
|
| 17 |
+
"processor_class": "SiglipProcessor",
|
| 18 |
+
"resample": 2,
|
| 19 |
+
"rescale_factor": 0.00392156862745098,
|
| 20 |
+
"size": {
|
| 21 |
+
"height": 224,
|
| 22 |
+
"width": 224
|
| 23 |
+
}
|
| 24 |
+
}
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eff7895602b654dfe6fc27f9bfd70976ed586448b1d172d1abb6cf2d3bccf43
|
| 3 |
+
size 5304
|