Upload folder using huggingface_hub
Browse files- checkpoint-1392/config.json +294 -0
- checkpoint-1392/model.safetensors +3 -0
- checkpoint-1392/optimizer.pt +3 -0
- checkpoint-1392/preprocessor_config.json +24 -0
- checkpoint-1392/rng_state.pth +3 -0
- checkpoint-1392/scheduler.pt +3 -0
- checkpoint-1392/trainer_state.json +78 -0
- checkpoint-1392/training_args.bin +3 -0
- checkpoint-1856/config.json +294 -0
- checkpoint-1856/model.safetensors +3 -0
- checkpoint-1856/optimizer.pt +3 -0
- checkpoint-1856/preprocessor_config.json +24 -0
- checkpoint-1856/rng_state.pth +3 -0
- checkpoint-1856/scheduler.pt +3 -0
- checkpoint-1856/trainer_state.json +95 -0
- checkpoint-1856/training_args.bin +3 -0
- checkpoint-2320/config.json +294 -0
- checkpoint-2320/model.safetensors +3 -0
- checkpoint-2320/optimizer.pt +3 -0
- checkpoint-2320/preprocessor_config.json +24 -0
- checkpoint-2320/rng_state.pth +3 -0
- checkpoint-2320/scheduler.pt +3 -0
- checkpoint-2320/trainer_state.json +112 -0
- checkpoint-2320/training_args.bin +3 -0
- checkpoint-464/config.json +294 -0
- checkpoint-464/model.safetensors +3 -0
- checkpoint-464/optimizer.pt +3 -0
- checkpoint-464/preprocessor_config.json +24 -0
- checkpoint-464/rng_state.pth +3 -0
- checkpoint-464/scheduler.pt +3 -0
- checkpoint-464/trainer_state.json +44 -0
- checkpoint-464/training_args.bin +3 -0
- checkpoint-928/config.json +294 -0
- checkpoint-928/model.safetensors +3 -0
- checkpoint-928/optimizer.pt +3 -0
- checkpoint-928/preprocessor_config.json +24 -0
- checkpoint-928/rng_state.pth +3 -0
- checkpoint-928/scheduler.pt +3 -0
- checkpoint-928/trainer_state.json +61 -0
- checkpoint-928/training_args.bin +3 -0
- config.json +294 -0
- model.safetensors +3 -0
- preprocessor_config.json +24 -0
- training_args.bin +3 -0
checkpoint-1392/config.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "aircraft_carrier",
|
| 7 |
+
"1": "alarm_clock",
|
| 8 |
+
"2": "ant",
|
| 9 |
+
"3": "anvil",
|
| 10 |
+
"4": "asparagus",
|
| 11 |
+
"5": "axe",
|
| 12 |
+
"6": "banana",
|
| 13 |
+
"7": "basket",
|
| 14 |
+
"8": "bathtub",
|
| 15 |
+
"9": "bear",
|
| 16 |
+
"10": "bee",
|
| 17 |
+
"11": "bird",
|
| 18 |
+
"12": "blackberry",
|
| 19 |
+
"13": "blueberry",
|
| 20 |
+
"14": "bottlecap",
|
| 21 |
+
"15": "broccoli",
|
| 22 |
+
"16": "bus",
|
| 23 |
+
"17": "butterfly",
|
| 24 |
+
"18": "cactus",
|
| 25 |
+
"19": "cake",
|
| 26 |
+
"20": "calculator",
|
| 27 |
+
"21": "camel",
|
| 28 |
+
"22": "camera",
|
| 29 |
+
"23": "candle",
|
| 30 |
+
"24": "cannon",
|
| 31 |
+
"25": "canoe",
|
| 32 |
+
"26": "carrot",
|
| 33 |
+
"27": "castle",
|
| 34 |
+
"28": "cat",
|
| 35 |
+
"29": "ceiling_fan",
|
| 36 |
+
"30": "cell_phone",
|
| 37 |
+
"31": "cello",
|
| 38 |
+
"32": "chair",
|
| 39 |
+
"33": "chandelier",
|
| 40 |
+
"34": "coffee_cup",
|
| 41 |
+
"35": "compass",
|
| 42 |
+
"36": "computer",
|
| 43 |
+
"37": "cow",
|
| 44 |
+
"38": "crab",
|
| 45 |
+
"39": "crocodile",
|
| 46 |
+
"40": "cruise_ship",
|
| 47 |
+
"41": "dog",
|
| 48 |
+
"42": "dolphin",
|
| 49 |
+
"43": "dragon",
|
| 50 |
+
"44": "drums",
|
| 51 |
+
"45": "duck",
|
| 52 |
+
"46": "dumbbell",
|
| 53 |
+
"47": "elephant",
|
| 54 |
+
"48": "eyeglasses",
|
| 55 |
+
"49": "feather",
|
| 56 |
+
"50": "fence",
|
| 57 |
+
"51": "fish",
|
| 58 |
+
"52": "flamingo",
|
| 59 |
+
"53": "flower",
|
| 60 |
+
"54": "foot",
|
| 61 |
+
"55": "fork",
|
| 62 |
+
"56": "frog",
|
| 63 |
+
"57": "giraffe",
|
| 64 |
+
"58": "goatee",
|
| 65 |
+
"59": "grapes",
|
| 66 |
+
"60": "guitar",
|
| 67 |
+
"61": "hammer",
|
| 68 |
+
"62": "helicopter",
|
| 69 |
+
"63": "helmet",
|
| 70 |
+
"64": "horse",
|
| 71 |
+
"65": "kangaroo",
|
| 72 |
+
"66": "lantern",
|
| 73 |
+
"67": "laptop",
|
| 74 |
+
"68": "leaf",
|
| 75 |
+
"69": "lion",
|
| 76 |
+
"70": "lipstick",
|
| 77 |
+
"71": "lobster",
|
| 78 |
+
"72": "microphone",
|
| 79 |
+
"73": "monkey",
|
| 80 |
+
"74": "mosquito",
|
| 81 |
+
"75": "mouse",
|
| 82 |
+
"76": "mug",
|
| 83 |
+
"77": "mushroom",
|
| 84 |
+
"78": "onion",
|
| 85 |
+
"79": "panda",
|
| 86 |
+
"80": "peanut",
|
| 87 |
+
"81": "pear",
|
| 88 |
+
"82": "peas",
|
| 89 |
+
"83": "pencil",
|
| 90 |
+
"84": "penguin",
|
| 91 |
+
"85": "pig",
|
| 92 |
+
"86": "pillow",
|
| 93 |
+
"87": "pineapple",
|
| 94 |
+
"88": "potato",
|
| 95 |
+
"89": "power_outlet",
|
| 96 |
+
"90": "purse",
|
| 97 |
+
"91": "rabbit",
|
| 98 |
+
"92": "raccoon",
|
| 99 |
+
"93": "rhinoceros",
|
| 100 |
+
"94": "rifle",
|
| 101 |
+
"95": "saxophone",
|
| 102 |
+
"96": "screwdriver",
|
| 103 |
+
"97": "sea_turtle",
|
| 104 |
+
"98": "see_saw",
|
| 105 |
+
"99": "sheep",
|
| 106 |
+
"100": "shoe",
|
| 107 |
+
"101": "skateboard",
|
| 108 |
+
"102": "snake",
|
| 109 |
+
"103": "speedboat",
|
| 110 |
+
"104": "spider",
|
| 111 |
+
"105": "squirrel",
|
| 112 |
+
"106": "strawberry",
|
| 113 |
+
"107": "streetlight",
|
| 114 |
+
"108": "string_bean",
|
| 115 |
+
"109": "submarine",
|
| 116 |
+
"110": "swan",
|
| 117 |
+
"111": "table",
|
| 118 |
+
"112": "teapot",
|
| 119 |
+
"113": "teddy-bear",
|
| 120 |
+
"114": "television",
|
| 121 |
+
"115": "the_Eiffel_Tower",
|
| 122 |
+
"116": "the_Great_Wall_of_China",
|
| 123 |
+
"117": "tiger",
|
| 124 |
+
"118": "toe",
|
| 125 |
+
"119": "train",
|
| 126 |
+
"120": "truck",
|
| 127 |
+
"121": "umbrella",
|
| 128 |
+
"122": "vase",
|
| 129 |
+
"123": "watermelon",
|
| 130 |
+
"124": "whale",
|
| 131 |
+
"125": "zebra"
|
| 132 |
+
},
|
| 133 |
+
"initializer_factor": 1.0,
|
| 134 |
+
"label2id": {
|
| 135 |
+
"aircraft_carrier": 0,
|
| 136 |
+
"alarm_clock": 1,
|
| 137 |
+
"ant": 2,
|
| 138 |
+
"anvil": 3,
|
| 139 |
+
"asparagus": 4,
|
| 140 |
+
"axe": 5,
|
| 141 |
+
"banana": 6,
|
| 142 |
+
"basket": 7,
|
| 143 |
+
"bathtub": 8,
|
| 144 |
+
"bear": 9,
|
| 145 |
+
"bee": 10,
|
| 146 |
+
"bird": 11,
|
| 147 |
+
"blackberry": 12,
|
| 148 |
+
"blueberry": 13,
|
| 149 |
+
"bottlecap": 14,
|
| 150 |
+
"broccoli": 15,
|
| 151 |
+
"bus": 16,
|
| 152 |
+
"butterfly": 17,
|
| 153 |
+
"cactus": 18,
|
| 154 |
+
"cake": 19,
|
| 155 |
+
"calculator": 20,
|
| 156 |
+
"camel": 21,
|
| 157 |
+
"camera": 22,
|
| 158 |
+
"candle": 23,
|
| 159 |
+
"cannon": 24,
|
| 160 |
+
"canoe": 25,
|
| 161 |
+
"carrot": 26,
|
| 162 |
+
"castle": 27,
|
| 163 |
+
"cat": 28,
|
| 164 |
+
"ceiling_fan": 29,
|
| 165 |
+
"cell_phone": 30,
|
| 166 |
+
"cello": 31,
|
| 167 |
+
"chair": 32,
|
| 168 |
+
"chandelier": 33,
|
| 169 |
+
"coffee_cup": 34,
|
| 170 |
+
"compass": 35,
|
| 171 |
+
"computer": 36,
|
| 172 |
+
"cow": 37,
|
| 173 |
+
"crab": 38,
|
| 174 |
+
"crocodile": 39,
|
| 175 |
+
"cruise_ship": 40,
|
| 176 |
+
"dog": 41,
|
| 177 |
+
"dolphin": 42,
|
| 178 |
+
"dragon": 43,
|
| 179 |
+
"drums": 44,
|
| 180 |
+
"duck": 45,
|
| 181 |
+
"dumbbell": 46,
|
| 182 |
+
"elephant": 47,
|
| 183 |
+
"eyeglasses": 48,
|
| 184 |
+
"feather": 49,
|
| 185 |
+
"fence": 50,
|
| 186 |
+
"fish": 51,
|
| 187 |
+
"flamingo": 52,
|
| 188 |
+
"flower": 53,
|
| 189 |
+
"foot": 54,
|
| 190 |
+
"fork": 55,
|
| 191 |
+
"frog": 56,
|
| 192 |
+
"giraffe": 57,
|
| 193 |
+
"goatee": 58,
|
| 194 |
+
"grapes": 59,
|
| 195 |
+
"guitar": 60,
|
| 196 |
+
"hammer": 61,
|
| 197 |
+
"helicopter": 62,
|
| 198 |
+
"helmet": 63,
|
| 199 |
+
"horse": 64,
|
| 200 |
+
"kangaroo": 65,
|
| 201 |
+
"lantern": 66,
|
| 202 |
+
"laptop": 67,
|
| 203 |
+
"leaf": 68,
|
| 204 |
+
"lion": 69,
|
| 205 |
+
"lipstick": 70,
|
| 206 |
+
"lobster": 71,
|
| 207 |
+
"microphone": 72,
|
| 208 |
+
"monkey": 73,
|
| 209 |
+
"mosquito": 74,
|
| 210 |
+
"mouse": 75,
|
| 211 |
+
"mug": 76,
|
| 212 |
+
"mushroom": 77,
|
| 213 |
+
"onion": 78,
|
| 214 |
+
"panda": 79,
|
| 215 |
+
"peanut": 80,
|
| 216 |
+
"pear": 81,
|
| 217 |
+
"peas": 82,
|
| 218 |
+
"pencil": 83,
|
| 219 |
+
"penguin": 84,
|
| 220 |
+
"pig": 85,
|
| 221 |
+
"pillow": 86,
|
| 222 |
+
"pineapple": 87,
|
| 223 |
+
"potato": 88,
|
| 224 |
+
"power_outlet": 89,
|
| 225 |
+
"purse": 90,
|
| 226 |
+
"rabbit": 91,
|
| 227 |
+
"raccoon": 92,
|
| 228 |
+
"rhinoceros": 93,
|
| 229 |
+
"rifle": 94,
|
| 230 |
+
"saxophone": 95,
|
| 231 |
+
"screwdriver": 96,
|
| 232 |
+
"sea_turtle": 97,
|
| 233 |
+
"see_saw": 98,
|
| 234 |
+
"sheep": 99,
|
| 235 |
+
"shoe": 100,
|
| 236 |
+
"skateboard": 101,
|
| 237 |
+
"snake": 102,
|
| 238 |
+
"speedboat": 103,
|
| 239 |
+
"spider": 104,
|
| 240 |
+
"squirrel": 105,
|
| 241 |
+
"strawberry": 106,
|
| 242 |
+
"streetlight": 107,
|
| 243 |
+
"string_bean": 108,
|
| 244 |
+
"submarine": 109,
|
| 245 |
+
"swan": 110,
|
| 246 |
+
"table": 111,
|
| 247 |
+
"teapot": 112,
|
| 248 |
+
"teddy-bear": 113,
|
| 249 |
+
"television": 114,
|
| 250 |
+
"the_Eiffel_Tower": 115,
|
| 251 |
+
"the_Great_Wall_of_China": 116,
|
| 252 |
+
"tiger": 117,
|
| 253 |
+
"toe": 118,
|
| 254 |
+
"train": 119,
|
| 255 |
+
"truck": 120,
|
| 256 |
+
"umbrella": 121,
|
| 257 |
+
"vase": 122,
|
| 258 |
+
"watermelon": 123,
|
| 259 |
+
"whale": 124,
|
| 260 |
+
"zebra": 125
|
| 261 |
+
},
|
| 262 |
+
"model_type": "siglip",
|
| 263 |
+
"problem_type": "single_label_classification",
|
| 264 |
+
"text_config": {
|
| 265 |
+
"attention_dropout": 0.0,
|
| 266 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 267 |
+
"hidden_size": 768,
|
| 268 |
+
"intermediate_size": 3072,
|
| 269 |
+
"layer_norm_eps": 1e-06,
|
| 270 |
+
"max_position_embeddings": 64,
|
| 271 |
+
"model_type": "siglip_text_model",
|
| 272 |
+
"num_attention_heads": 12,
|
| 273 |
+
"num_hidden_layers": 12,
|
| 274 |
+
"projection_size": 768,
|
| 275 |
+
"torch_dtype": "float32",
|
| 276 |
+
"vocab_size": 256000
|
| 277 |
+
},
|
| 278 |
+
"torch_dtype": "float32",
|
| 279 |
+
"transformers_version": "4.51.0.dev0",
|
| 280 |
+
"vision_config": {
|
| 281 |
+
"attention_dropout": 0.0,
|
| 282 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 283 |
+
"hidden_size": 768,
|
| 284 |
+
"image_size": 224,
|
| 285 |
+
"intermediate_size": 3072,
|
| 286 |
+
"layer_norm_eps": 1e-06,
|
| 287 |
+
"model_type": "siglip_vision_model",
|
| 288 |
+
"num_attention_heads": 12,
|
| 289 |
+
"num_channels": 3,
|
| 290 |
+
"num_hidden_layers": 12,
|
| 291 |
+
"patch_size": 16,
|
| 292 |
+
"torch_dtype": "float32"
|
| 293 |
+
}
|
| 294 |
+
}
|
checkpoint-1392/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6d0934ecc09ccc45a8753d8992a8cdad9782a255bfa500c04d449dc4135aa926
|
| 3 |
+
size 371949432
|
checkpoint-1392/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:21ab779450cf11177367de31310572d7098fb6f7025419492e4d466892a108d0
|
| 3 |
+
size 687318522
|
checkpoint-1392/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-1392/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a25adcb0b76226196097ddae86610355831c1ca39dfb917f1cb923de385ba7f4
|
| 3 |
+
size 14244
|
checkpoint-1392/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:70561b3085430b7ae1ed4f8c7d5c8e77d43533aa11d0b69a4415c4f1a81149e3
|
| 3 |
+
size 1064
|
checkpoint-1392/trainer_state.json
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 1392,
|
| 3 |
+
"best_metric": 0.7880541086196899,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-1392",
|
| 5 |
+
"epoch": 3.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 1392,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 1.0,
|
| 14 |
+
"eval_accuracy": 0.6359832635983264,
|
| 15 |
+
"eval_loss": 2.0701308250427246,
|
| 16 |
+
"eval_model_preparation_time": 0.0031,
|
| 17 |
+
"eval_runtime": 215.6864,
|
| 18 |
+
"eval_samples_per_second": 68.702,
|
| 19 |
+
"eval_steps_per_second": 8.591,
|
| 20 |
+
"step": 464
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 1.0775862068965518,
|
| 24 |
+
"grad_norm": 49.15087890625,
|
| 25 |
+
"learning_rate": 2.4052863436123347e-06,
|
| 26 |
+
"loss": 3.5107,
|
| 27 |
+
"step": 500
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 2.0,
|
| 31 |
+
"eval_accuracy": 0.8090160615467674,
|
| 32 |
+
"eval_loss": 1.0770337581634521,
|
| 33 |
+
"eval_model_preparation_time": 0.0031,
|
| 34 |
+
"eval_runtime": 213.0035,
|
| 35 |
+
"eval_samples_per_second": 69.567,
|
| 36 |
+
"eval_steps_per_second": 8.699,
|
| 37 |
+
"step": 928
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 2.1551724137931036,
|
| 41 |
+
"grad_norm": 57.83103561401367,
|
| 42 |
+
"learning_rate": 1.7444933920704845e-06,
|
| 43 |
+
"loss": 1.3776,
|
| 44 |
+
"step": 1000
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 3.0,
|
| 48 |
+
"eval_accuracy": 0.8484275880685652,
|
| 49 |
+
"eval_loss": 0.7880541086196899,
|
| 50 |
+
"eval_model_preparation_time": 0.0031,
|
| 51 |
+
"eval_runtime": 213.1605,
|
| 52 |
+
"eval_samples_per_second": 69.516,
|
| 53 |
+
"eval_steps_per_second": 8.693,
|
| 54 |
+
"step": 1392
|
| 55 |
+
}
|
| 56 |
+
],
|
| 57 |
+
"logging_steps": 500,
|
| 58 |
+
"max_steps": 2320,
|
| 59 |
+
"num_input_tokens_seen": 0,
|
| 60 |
+
"num_train_epochs": 5,
|
| 61 |
+
"save_steps": 500,
|
| 62 |
+
"stateful_callbacks": {
|
| 63 |
+
"TrainerControl": {
|
| 64 |
+
"args": {
|
| 65 |
+
"should_epoch_stop": false,
|
| 66 |
+
"should_evaluate": false,
|
| 67 |
+
"should_log": false,
|
| 68 |
+
"should_save": true,
|
| 69 |
+
"should_training_stop": false
|
| 70 |
+
},
|
| 71 |
+
"attributes": {}
|
| 72 |
+
}
|
| 73 |
+
},
|
| 74 |
+
"total_flos": 3.7270953003136205e+18,
|
| 75 |
+
"train_batch_size": 32,
|
| 76 |
+
"trial_name": null,
|
| 77 |
+
"trial_params": null
|
| 78 |
+
}
|
checkpoint-1392/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d60dfc3a7f3feaaf5fab64c05e89c94efd07d977dc0d1cf7a28f5e86e8cbf7e5
|
| 3 |
+
size 5304
|
checkpoint-1856/config.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "aircraft_carrier",
|
| 7 |
+
"1": "alarm_clock",
|
| 8 |
+
"2": "ant",
|
| 9 |
+
"3": "anvil",
|
| 10 |
+
"4": "asparagus",
|
| 11 |
+
"5": "axe",
|
| 12 |
+
"6": "banana",
|
| 13 |
+
"7": "basket",
|
| 14 |
+
"8": "bathtub",
|
| 15 |
+
"9": "bear",
|
| 16 |
+
"10": "bee",
|
| 17 |
+
"11": "bird",
|
| 18 |
+
"12": "blackberry",
|
| 19 |
+
"13": "blueberry",
|
| 20 |
+
"14": "bottlecap",
|
| 21 |
+
"15": "broccoli",
|
| 22 |
+
"16": "bus",
|
| 23 |
+
"17": "butterfly",
|
| 24 |
+
"18": "cactus",
|
| 25 |
+
"19": "cake",
|
| 26 |
+
"20": "calculator",
|
| 27 |
+
"21": "camel",
|
| 28 |
+
"22": "camera",
|
| 29 |
+
"23": "candle",
|
| 30 |
+
"24": "cannon",
|
| 31 |
+
"25": "canoe",
|
| 32 |
+
"26": "carrot",
|
| 33 |
+
"27": "castle",
|
| 34 |
+
"28": "cat",
|
| 35 |
+
"29": "ceiling_fan",
|
| 36 |
+
"30": "cell_phone",
|
| 37 |
+
"31": "cello",
|
| 38 |
+
"32": "chair",
|
| 39 |
+
"33": "chandelier",
|
| 40 |
+
"34": "coffee_cup",
|
| 41 |
+
"35": "compass",
|
| 42 |
+
"36": "computer",
|
| 43 |
+
"37": "cow",
|
| 44 |
+
"38": "crab",
|
| 45 |
+
"39": "crocodile",
|
| 46 |
+
"40": "cruise_ship",
|
| 47 |
+
"41": "dog",
|
| 48 |
+
"42": "dolphin",
|
| 49 |
+
"43": "dragon",
|
| 50 |
+
"44": "drums",
|
| 51 |
+
"45": "duck",
|
| 52 |
+
"46": "dumbbell",
|
| 53 |
+
"47": "elephant",
|
| 54 |
+
"48": "eyeglasses",
|
| 55 |
+
"49": "feather",
|
| 56 |
+
"50": "fence",
|
| 57 |
+
"51": "fish",
|
| 58 |
+
"52": "flamingo",
|
| 59 |
+
"53": "flower",
|
| 60 |
+
"54": "foot",
|
| 61 |
+
"55": "fork",
|
| 62 |
+
"56": "frog",
|
| 63 |
+
"57": "giraffe",
|
| 64 |
+
"58": "goatee",
|
| 65 |
+
"59": "grapes",
|
| 66 |
+
"60": "guitar",
|
| 67 |
+
"61": "hammer",
|
| 68 |
+
"62": "helicopter",
|
| 69 |
+
"63": "helmet",
|
| 70 |
+
"64": "horse",
|
| 71 |
+
"65": "kangaroo",
|
| 72 |
+
"66": "lantern",
|
| 73 |
+
"67": "laptop",
|
| 74 |
+
"68": "leaf",
|
| 75 |
+
"69": "lion",
|
| 76 |
+
"70": "lipstick",
|
| 77 |
+
"71": "lobster",
|
| 78 |
+
"72": "microphone",
|
| 79 |
+
"73": "monkey",
|
| 80 |
+
"74": "mosquito",
|
| 81 |
+
"75": "mouse",
|
| 82 |
+
"76": "mug",
|
| 83 |
+
"77": "mushroom",
|
| 84 |
+
"78": "onion",
|
| 85 |
+
"79": "panda",
|
| 86 |
+
"80": "peanut",
|
| 87 |
+
"81": "pear",
|
| 88 |
+
"82": "peas",
|
| 89 |
+
"83": "pencil",
|
| 90 |
+
"84": "penguin",
|
| 91 |
+
"85": "pig",
|
| 92 |
+
"86": "pillow",
|
| 93 |
+
"87": "pineapple",
|
| 94 |
+
"88": "potato",
|
| 95 |
+
"89": "power_outlet",
|
| 96 |
+
"90": "purse",
|
| 97 |
+
"91": "rabbit",
|
| 98 |
+
"92": "raccoon",
|
| 99 |
+
"93": "rhinoceros",
|
| 100 |
+
"94": "rifle",
|
| 101 |
+
"95": "saxophone",
|
| 102 |
+
"96": "screwdriver",
|
| 103 |
+
"97": "sea_turtle",
|
| 104 |
+
"98": "see_saw",
|
| 105 |
+
"99": "sheep",
|
| 106 |
+
"100": "shoe",
|
| 107 |
+
"101": "skateboard",
|
| 108 |
+
"102": "snake",
|
| 109 |
+
"103": "speedboat",
|
| 110 |
+
"104": "spider",
|
| 111 |
+
"105": "squirrel",
|
| 112 |
+
"106": "strawberry",
|
| 113 |
+
"107": "streetlight",
|
| 114 |
+
"108": "string_bean",
|
| 115 |
+
"109": "submarine",
|
| 116 |
+
"110": "swan",
|
| 117 |
+
"111": "table",
|
| 118 |
+
"112": "teapot",
|
| 119 |
+
"113": "teddy-bear",
|
| 120 |
+
"114": "television",
|
| 121 |
+
"115": "the_Eiffel_Tower",
|
| 122 |
+
"116": "the_Great_Wall_of_China",
|
| 123 |
+
"117": "tiger",
|
| 124 |
+
"118": "toe",
|
| 125 |
+
"119": "train",
|
| 126 |
+
"120": "truck",
|
| 127 |
+
"121": "umbrella",
|
| 128 |
+
"122": "vase",
|
| 129 |
+
"123": "watermelon",
|
| 130 |
+
"124": "whale",
|
| 131 |
+
"125": "zebra"
|
| 132 |
+
},
|
| 133 |
+
"initializer_factor": 1.0,
|
| 134 |
+
"label2id": {
|
| 135 |
+
"aircraft_carrier": 0,
|
| 136 |
+
"alarm_clock": 1,
|
| 137 |
+
"ant": 2,
|
| 138 |
+
"anvil": 3,
|
| 139 |
+
"asparagus": 4,
|
| 140 |
+
"axe": 5,
|
| 141 |
+
"banana": 6,
|
| 142 |
+
"basket": 7,
|
| 143 |
+
"bathtub": 8,
|
| 144 |
+
"bear": 9,
|
| 145 |
+
"bee": 10,
|
| 146 |
+
"bird": 11,
|
| 147 |
+
"blackberry": 12,
|
| 148 |
+
"blueberry": 13,
|
| 149 |
+
"bottlecap": 14,
|
| 150 |
+
"broccoli": 15,
|
| 151 |
+
"bus": 16,
|
| 152 |
+
"butterfly": 17,
|
| 153 |
+
"cactus": 18,
|
| 154 |
+
"cake": 19,
|
| 155 |
+
"calculator": 20,
|
| 156 |
+
"camel": 21,
|
| 157 |
+
"camera": 22,
|
| 158 |
+
"candle": 23,
|
| 159 |
+
"cannon": 24,
|
| 160 |
+
"canoe": 25,
|
| 161 |
+
"carrot": 26,
|
| 162 |
+
"castle": 27,
|
| 163 |
+
"cat": 28,
|
| 164 |
+
"ceiling_fan": 29,
|
| 165 |
+
"cell_phone": 30,
|
| 166 |
+
"cello": 31,
|
| 167 |
+
"chair": 32,
|
| 168 |
+
"chandelier": 33,
|
| 169 |
+
"coffee_cup": 34,
|
| 170 |
+
"compass": 35,
|
| 171 |
+
"computer": 36,
|
| 172 |
+
"cow": 37,
|
| 173 |
+
"crab": 38,
|
| 174 |
+
"crocodile": 39,
|
| 175 |
+
"cruise_ship": 40,
|
| 176 |
+
"dog": 41,
|
| 177 |
+
"dolphin": 42,
|
| 178 |
+
"dragon": 43,
|
| 179 |
+
"drums": 44,
|
| 180 |
+
"duck": 45,
|
| 181 |
+
"dumbbell": 46,
|
| 182 |
+
"elephant": 47,
|
| 183 |
+
"eyeglasses": 48,
|
| 184 |
+
"feather": 49,
|
| 185 |
+
"fence": 50,
|
| 186 |
+
"fish": 51,
|
| 187 |
+
"flamingo": 52,
|
| 188 |
+
"flower": 53,
|
| 189 |
+
"foot": 54,
|
| 190 |
+
"fork": 55,
|
| 191 |
+
"frog": 56,
|
| 192 |
+
"giraffe": 57,
|
| 193 |
+
"goatee": 58,
|
| 194 |
+
"grapes": 59,
|
| 195 |
+
"guitar": 60,
|
| 196 |
+
"hammer": 61,
|
| 197 |
+
"helicopter": 62,
|
| 198 |
+
"helmet": 63,
|
| 199 |
+
"horse": 64,
|
| 200 |
+
"kangaroo": 65,
|
| 201 |
+
"lantern": 66,
|
| 202 |
+
"laptop": 67,
|
| 203 |
+
"leaf": 68,
|
| 204 |
+
"lion": 69,
|
| 205 |
+
"lipstick": 70,
|
| 206 |
+
"lobster": 71,
|
| 207 |
+
"microphone": 72,
|
| 208 |
+
"monkey": 73,
|
| 209 |
+
"mosquito": 74,
|
| 210 |
+
"mouse": 75,
|
| 211 |
+
"mug": 76,
|
| 212 |
+
"mushroom": 77,
|
| 213 |
+
"onion": 78,
|
| 214 |
+
"panda": 79,
|
| 215 |
+
"peanut": 80,
|
| 216 |
+
"pear": 81,
|
| 217 |
+
"peas": 82,
|
| 218 |
+
"pencil": 83,
|
| 219 |
+
"penguin": 84,
|
| 220 |
+
"pig": 85,
|
| 221 |
+
"pillow": 86,
|
| 222 |
+
"pineapple": 87,
|
| 223 |
+
"potato": 88,
|
| 224 |
+
"power_outlet": 89,
|
| 225 |
+
"purse": 90,
|
| 226 |
+
"rabbit": 91,
|
| 227 |
+
"raccoon": 92,
|
| 228 |
+
"rhinoceros": 93,
|
| 229 |
+
"rifle": 94,
|
| 230 |
+
"saxophone": 95,
|
| 231 |
+
"screwdriver": 96,
|
| 232 |
+
"sea_turtle": 97,
|
| 233 |
+
"see_saw": 98,
|
| 234 |
+
"sheep": 99,
|
| 235 |
+
"shoe": 100,
|
| 236 |
+
"skateboard": 101,
|
| 237 |
+
"snake": 102,
|
| 238 |
+
"speedboat": 103,
|
| 239 |
+
"spider": 104,
|
| 240 |
+
"squirrel": 105,
|
| 241 |
+
"strawberry": 106,
|
| 242 |
+
"streetlight": 107,
|
| 243 |
+
"string_bean": 108,
|
| 244 |
+
"submarine": 109,
|
| 245 |
+
"swan": 110,
|
| 246 |
+
"table": 111,
|
| 247 |
+
"teapot": 112,
|
| 248 |
+
"teddy-bear": 113,
|
| 249 |
+
"television": 114,
|
| 250 |
+
"the_Eiffel_Tower": 115,
|
| 251 |
+
"the_Great_Wall_of_China": 116,
|
| 252 |
+
"tiger": 117,
|
| 253 |
+
"toe": 118,
|
| 254 |
+
"train": 119,
|
| 255 |
+
"truck": 120,
|
| 256 |
+
"umbrella": 121,
|
| 257 |
+
"vase": 122,
|
| 258 |
+
"watermelon": 123,
|
| 259 |
+
"whale": 124,
|
| 260 |
+
"zebra": 125
|
| 261 |
+
},
|
| 262 |
+
"model_type": "siglip",
|
| 263 |
+
"problem_type": "single_label_classification",
|
| 264 |
+
"text_config": {
|
| 265 |
+
"attention_dropout": 0.0,
|
| 266 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 267 |
+
"hidden_size": 768,
|
| 268 |
+
"intermediate_size": 3072,
|
| 269 |
+
"layer_norm_eps": 1e-06,
|
| 270 |
+
"max_position_embeddings": 64,
|
| 271 |
+
"model_type": "siglip_text_model",
|
| 272 |
+
"num_attention_heads": 12,
|
| 273 |
+
"num_hidden_layers": 12,
|
| 274 |
+
"projection_size": 768,
|
| 275 |
+
"torch_dtype": "float32",
|
| 276 |
+
"vocab_size": 256000
|
| 277 |
+
},
|
| 278 |
+
"torch_dtype": "float32",
|
| 279 |
+
"transformers_version": "4.51.0.dev0",
|
| 280 |
+
"vision_config": {
|
| 281 |
+
"attention_dropout": 0.0,
|
| 282 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 283 |
+
"hidden_size": 768,
|
| 284 |
+
"image_size": 224,
|
| 285 |
+
"intermediate_size": 3072,
|
| 286 |
+
"layer_norm_eps": 1e-06,
|
| 287 |
+
"model_type": "siglip_vision_model",
|
| 288 |
+
"num_attention_heads": 12,
|
| 289 |
+
"num_channels": 3,
|
| 290 |
+
"num_hidden_layers": 12,
|
| 291 |
+
"patch_size": 16,
|
| 292 |
+
"torch_dtype": "float32"
|
| 293 |
+
}
|
| 294 |
+
}
|
checkpoint-1856/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:98c75a63b2d442e400590214a8a78cfad063f59b5b0d992ffc68c5a07b5ec7cb
|
| 3 |
+
size 371949432
|
checkpoint-1856/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dbd5baeac158f5a7420d3362e578cb8acee2e8ac94c1341f23ab294ef9dd447a
|
| 3 |
+
size 687318522
|
checkpoint-1856/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-1856/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:755f786b87a2759da449e45144b0cea263dc29e45c73bfe9a5ef7721d95bfda6
|
| 3 |
+
size 14244
|
checkpoint-1856/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:649186c849e26de120344d7ca3e7bac468a9ca99f83af7618583289fa1189c1b
|
| 3 |
+
size 1064
|
checkpoint-1856/trainer_state.json
ADDED
|
@@ -0,0 +1,95 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 1856,
|
| 3 |
+
"best_metric": 0.6761099696159363,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-1856",
|
| 5 |
+
"epoch": 4.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 1856,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 1.0,
|
| 14 |
+
"eval_accuracy": 0.6359832635983264,
|
| 15 |
+
"eval_loss": 2.0701308250427246,
|
| 16 |
+
"eval_model_preparation_time": 0.0031,
|
| 17 |
+
"eval_runtime": 215.6864,
|
| 18 |
+
"eval_samples_per_second": 68.702,
|
| 19 |
+
"eval_steps_per_second": 8.591,
|
| 20 |
+
"step": 464
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 1.0775862068965518,
|
| 24 |
+
"grad_norm": 49.15087890625,
|
| 25 |
+
"learning_rate": 2.4052863436123347e-06,
|
| 26 |
+
"loss": 3.5107,
|
| 27 |
+
"step": 500
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 2.0,
|
| 31 |
+
"eval_accuracy": 0.8090160615467674,
|
| 32 |
+
"eval_loss": 1.0770337581634521,
|
| 33 |
+
"eval_model_preparation_time": 0.0031,
|
| 34 |
+
"eval_runtime": 213.0035,
|
| 35 |
+
"eval_samples_per_second": 69.567,
|
| 36 |
+
"eval_steps_per_second": 8.699,
|
| 37 |
+
"step": 928
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 2.1551724137931036,
|
| 41 |
+
"grad_norm": 57.83103561401367,
|
| 42 |
+
"learning_rate": 1.7444933920704845e-06,
|
| 43 |
+
"loss": 1.3776,
|
| 44 |
+
"step": 1000
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 3.0,
|
| 48 |
+
"eval_accuracy": 0.8484275880685652,
|
| 49 |
+
"eval_loss": 0.7880541086196899,
|
| 50 |
+
"eval_model_preparation_time": 0.0031,
|
| 51 |
+
"eval_runtime": 213.1605,
|
| 52 |
+
"eval_samples_per_second": 69.516,
|
| 53 |
+
"eval_steps_per_second": 8.693,
|
| 54 |
+
"step": 1392
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 3.2327586206896552,
|
| 58 |
+
"grad_norm": 65.75613403320312,
|
| 59 |
+
"learning_rate": 1.0837004405286343e-06,
|
| 60 |
+
"loss": 0.8854,
|
| 61 |
+
"step": 1500
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"epoch": 4.0,
|
| 65 |
+
"eval_accuracy": 0.8636118234579565,
|
| 66 |
+
"eval_loss": 0.6761099696159363,
|
| 67 |
+
"eval_model_preparation_time": 0.0031,
|
| 68 |
+
"eval_runtime": 212.8345,
|
| 69 |
+
"eval_samples_per_second": 69.622,
|
| 70 |
+
"eval_steps_per_second": 8.706,
|
| 71 |
+
"step": 1856
|
| 72 |
+
}
|
| 73 |
+
],
|
| 74 |
+
"logging_steps": 500,
|
| 75 |
+
"max_steps": 2320,
|
| 76 |
+
"num_input_tokens_seen": 0,
|
| 77 |
+
"num_train_epochs": 5,
|
| 78 |
+
"save_steps": 500,
|
| 79 |
+
"stateful_callbacks": {
|
| 80 |
+
"TrainerControl": {
|
| 81 |
+
"args": {
|
| 82 |
+
"should_epoch_stop": false,
|
| 83 |
+
"should_evaluate": false,
|
| 84 |
+
"should_log": false,
|
| 85 |
+
"should_save": true,
|
| 86 |
+
"should_training_stop": false
|
| 87 |
+
},
|
| 88 |
+
"attributes": {}
|
| 89 |
+
}
|
| 90 |
+
},
|
| 91 |
+
"total_flos": 4.969460400418161e+18,
|
| 92 |
+
"train_batch_size": 32,
|
| 93 |
+
"trial_name": null,
|
| 94 |
+
"trial_params": null
|
| 95 |
+
}
|
checkpoint-1856/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d60dfc3a7f3feaaf5fab64c05e89c94efd07d977dc0d1cf7a28f5e86e8cbf7e5
|
| 3 |
+
size 5304
|
checkpoint-2320/config.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "aircraft_carrier",
|
| 7 |
+
"1": "alarm_clock",
|
| 8 |
+
"2": "ant",
|
| 9 |
+
"3": "anvil",
|
| 10 |
+
"4": "asparagus",
|
| 11 |
+
"5": "axe",
|
| 12 |
+
"6": "banana",
|
| 13 |
+
"7": "basket",
|
| 14 |
+
"8": "bathtub",
|
| 15 |
+
"9": "bear",
|
| 16 |
+
"10": "bee",
|
| 17 |
+
"11": "bird",
|
| 18 |
+
"12": "blackberry",
|
| 19 |
+
"13": "blueberry",
|
| 20 |
+
"14": "bottlecap",
|
| 21 |
+
"15": "broccoli",
|
| 22 |
+
"16": "bus",
|
| 23 |
+
"17": "butterfly",
|
| 24 |
+
"18": "cactus",
|
| 25 |
+
"19": "cake",
|
| 26 |
+
"20": "calculator",
|
| 27 |
+
"21": "camel",
|
| 28 |
+
"22": "camera",
|
| 29 |
+
"23": "candle",
|
| 30 |
+
"24": "cannon",
|
| 31 |
+
"25": "canoe",
|
| 32 |
+
"26": "carrot",
|
| 33 |
+
"27": "castle",
|
| 34 |
+
"28": "cat",
|
| 35 |
+
"29": "ceiling_fan",
|
| 36 |
+
"30": "cell_phone",
|
| 37 |
+
"31": "cello",
|
| 38 |
+
"32": "chair",
|
| 39 |
+
"33": "chandelier",
|
| 40 |
+
"34": "coffee_cup",
|
| 41 |
+
"35": "compass",
|
| 42 |
+
"36": "computer",
|
| 43 |
+
"37": "cow",
|
| 44 |
+
"38": "crab",
|
| 45 |
+
"39": "crocodile",
|
| 46 |
+
"40": "cruise_ship",
|
| 47 |
+
"41": "dog",
|
| 48 |
+
"42": "dolphin",
|
| 49 |
+
"43": "dragon",
|
| 50 |
+
"44": "drums",
|
| 51 |
+
"45": "duck",
|
| 52 |
+
"46": "dumbbell",
|
| 53 |
+
"47": "elephant",
|
| 54 |
+
"48": "eyeglasses",
|
| 55 |
+
"49": "feather",
|
| 56 |
+
"50": "fence",
|
| 57 |
+
"51": "fish",
|
| 58 |
+
"52": "flamingo",
|
| 59 |
+
"53": "flower",
|
| 60 |
+
"54": "foot",
|
| 61 |
+
"55": "fork",
|
| 62 |
+
"56": "frog",
|
| 63 |
+
"57": "giraffe",
|
| 64 |
+
"58": "goatee",
|
| 65 |
+
"59": "grapes",
|
| 66 |
+
"60": "guitar",
|
| 67 |
+
"61": "hammer",
|
| 68 |
+
"62": "helicopter",
|
| 69 |
+
"63": "helmet",
|
| 70 |
+
"64": "horse",
|
| 71 |
+
"65": "kangaroo",
|
| 72 |
+
"66": "lantern",
|
| 73 |
+
"67": "laptop",
|
| 74 |
+
"68": "leaf",
|
| 75 |
+
"69": "lion",
|
| 76 |
+
"70": "lipstick",
|
| 77 |
+
"71": "lobster",
|
| 78 |
+
"72": "microphone",
|
| 79 |
+
"73": "monkey",
|
| 80 |
+
"74": "mosquito",
|
| 81 |
+
"75": "mouse",
|
| 82 |
+
"76": "mug",
|
| 83 |
+
"77": "mushroom",
|
| 84 |
+
"78": "onion",
|
| 85 |
+
"79": "panda",
|
| 86 |
+
"80": "peanut",
|
| 87 |
+
"81": "pear",
|
| 88 |
+
"82": "peas",
|
| 89 |
+
"83": "pencil",
|
| 90 |
+
"84": "penguin",
|
| 91 |
+
"85": "pig",
|
| 92 |
+
"86": "pillow",
|
| 93 |
+
"87": "pineapple",
|
| 94 |
+
"88": "potato",
|
| 95 |
+
"89": "power_outlet",
|
| 96 |
+
"90": "purse",
|
| 97 |
+
"91": "rabbit",
|
| 98 |
+
"92": "raccoon",
|
| 99 |
+
"93": "rhinoceros",
|
| 100 |
+
"94": "rifle",
|
| 101 |
+
"95": "saxophone",
|
| 102 |
+
"96": "screwdriver",
|
| 103 |
+
"97": "sea_turtle",
|
| 104 |
+
"98": "see_saw",
|
| 105 |
+
"99": "sheep",
|
| 106 |
+
"100": "shoe",
|
| 107 |
+
"101": "skateboard",
|
| 108 |
+
"102": "snake",
|
| 109 |
+
"103": "speedboat",
|
| 110 |
+
"104": "spider",
|
| 111 |
+
"105": "squirrel",
|
| 112 |
+
"106": "strawberry",
|
| 113 |
+
"107": "streetlight",
|
| 114 |
+
"108": "string_bean",
|
| 115 |
+
"109": "submarine",
|
| 116 |
+
"110": "swan",
|
| 117 |
+
"111": "table",
|
| 118 |
+
"112": "teapot",
|
| 119 |
+
"113": "teddy-bear",
|
| 120 |
+
"114": "television",
|
| 121 |
+
"115": "the_Eiffel_Tower",
|
| 122 |
+
"116": "the_Great_Wall_of_China",
|
| 123 |
+
"117": "tiger",
|
| 124 |
+
"118": "toe",
|
| 125 |
+
"119": "train",
|
| 126 |
+
"120": "truck",
|
| 127 |
+
"121": "umbrella",
|
| 128 |
+
"122": "vase",
|
| 129 |
+
"123": "watermelon",
|
| 130 |
+
"124": "whale",
|
| 131 |
+
"125": "zebra"
|
| 132 |
+
},
|
| 133 |
+
"initializer_factor": 1.0,
|
| 134 |
+
"label2id": {
|
| 135 |
+
"aircraft_carrier": 0,
|
| 136 |
+
"alarm_clock": 1,
|
| 137 |
+
"ant": 2,
|
| 138 |
+
"anvil": 3,
|
| 139 |
+
"asparagus": 4,
|
| 140 |
+
"axe": 5,
|
| 141 |
+
"banana": 6,
|
| 142 |
+
"basket": 7,
|
| 143 |
+
"bathtub": 8,
|
| 144 |
+
"bear": 9,
|
| 145 |
+
"bee": 10,
|
| 146 |
+
"bird": 11,
|
| 147 |
+
"blackberry": 12,
|
| 148 |
+
"blueberry": 13,
|
| 149 |
+
"bottlecap": 14,
|
| 150 |
+
"broccoli": 15,
|
| 151 |
+
"bus": 16,
|
| 152 |
+
"butterfly": 17,
|
| 153 |
+
"cactus": 18,
|
| 154 |
+
"cake": 19,
|
| 155 |
+
"calculator": 20,
|
| 156 |
+
"camel": 21,
|
| 157 |
+
"camera": 22,
|
| 158 |
+
"candle": 23,
|
| 159 |
+
"cannon": 24,
|
| 160 |
+
"canoe": 25,
|
| 161 |
+
"carrot": 26,
|
| 162 |
+
"castle": 27,
|
| 163 |
+
"cat": 28,
|
| 164 |
+
"ceiling_fan": 29,
|
| 165 |
+
"cell_phone": 30,
|
| 166 |
+
"cello": 31,
|
| 167 |
+
"chair": 32,
|
| 168 |
+
"chandelier": 33,
|
| 169 |
+
"coffee_cup": 34,
|
| 170 |
+
"compass": 35,
|
| 171 |
+
"computer": 36,
|
| 172 |
+
"cow": 37,
|
| 173 |
+
"crab": 38,
|
| 174 |
+
"crocodile": 39,
|
| 175 |
+
"cruise_ship": 40,
|
| 176 |
+
"dog": 41,
|
| 177 |
+
"dolphin": 42,
|
| 178 |
+
"dragon": 43,
|
| 179 |
+
"drums": 44,
|
| 180 |
+
"duck": 45,
|
| 181 |
+
"dumbbell": 46,
|
| 182 |
+
"elephant": 47,
|
| 183 |
+
"eyeglasses": 48,
|
| 184 |
+
"feather": 49,
|
| 185 |
+
"fence": 50,
|
| 186 |
+
"fish": 51,
|
| 187 |
+
"flamingo": 52,
|
| 188 |
+
"flower": 53,
|
| 189 |
+
"foot": 54,
|
| 190 |
+
"fork": 55,
|
| 191 |
+
"frog": 56,
|
| 192 |
+
"giraffe": 57,
|
| 193 |
+
"goatee": 58,
|
| 194 |
+
"grapes": 59,
|
| 195 |
+
"guitar": 60,
|
| 196 |
+
"hammer": 61,
|
| 197 |
+
"helicopter": 62,
|
| 198 |
+
"helmet": 63,
|
| 199 |
+
"horse": 64,
|
| 200 |
+
"kangaroo": 65,
|
| 201 |
+
"lantern": 66,
|
| 202 |
+
"laptop": 67,
|
| 203 |
+
"leaf": 68,
|
| 204 |
+
"lion": 69,
|
| 205 |
+
"lipstick": 70,
|
| 206 |
+
"lobster": 71,
|
| 207 |
+
"microphone": 72,
|
| 208 |
+
"monkey": 73,
|
| 209 |
+
"mosquito": 74,
|
| 210 |
+
"mouse": 75,
|
| 211 |
+
"mug": 76,
|
| 212 |
+
"mushroom": 77,
|
| 213 |
+
"onion": 78,
|
| 214 |
+
"panda": 79,
|
| 215 |
+
"peanut": 80,
|
| 216 |
+
"pear": 81,
|
| 217 |
+
"peas": 82,
|
| 218 |
+
"pencil": 83,
|
| 219 |
+
"penguin": 84,
|
| 220 |
+
"pig": 85,
|
| 221 |
+
"pillow": 86,
|
| 222 |
+
"pineapple": 87,
|
| 223 |
+
"potato": 88,
|
| 224 |
+
"power_outlet": 89,
|
| 225 |
+
"purse": 90,
|
| 226 |
+
"rabbit": 91,
|
| 227 |
+
"raccoon": 92,
|
| 228 |
+
"rhinoceros": 93,
|
| 229 |
+
"rifle": 94,
|
| 230 |
+
"saxophone": 95,
|
| 231 |
+
"screwdriver": 96,
|
| 232 |
+
"sea_turtle": 97,
|
| 233 |
+
"see_saw": 98,
|
| 234 |
+
"sheep": 99,
|
| 235 |
+
"shoe": 100,
|
| 236 |
+
"skateboard": 101,
|
| 237 |
+
"snake": 102,
|
| 238 |
+
"speedboat": 103,
|
| 239 |
+
"spider": 104,
|
| 240 |
+
"squirrel": 105,
|
| 241 |
+
"strawberry": 106,
|
| 242 |
+
"streetlight": 107,
|
| 243 |
+
"string_bean": 108,
|
| 244 |
+
"submarine": 109,
|
| 245 |
+
"swan": 110,
|
| 246 |
+
"table": 111,
|
| 247 |
+
"teapot": 112,
|
| 248 |
+
"teddy-bear": 113,
|
| 249 |
+
"television": 114,
|
| 250 |
+
"the_Eiffel_Tower": 115,
|
| 251 |
+
"the_Great_Wall_of_China": 116,
|
| 252 |
+
"tiger": 117,
|
| 253 |
+
"toe": 118,
|
| 254 |
+
"train": 119,
|
| 255 |
+
"truck": 120,
|
| 256 |
+
"umbrella": 121,
|
| 257 |
+
"vase": 122,
|
| 258 |
+
"watermelon": 123,
|
| 259 |
+
"whale": 124,
|
| 260 |
+
"zebra": 125
|
| 261 |
+
},
|
| 262 |
+
"model_type": "siglip",
|
| 263 |
+
"problem_type": "single_label_classification",
|
| 264 |
+
"text_config": {
|
| 265 |
+
"attention_dropout": 0.0,
|
| 266 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 267 |
+
"hidden_size": 768,
|
| 268 |
+
"intermediate_size": 3072,
|
| 269 |
+
"layer_norm_eps": 1e-06,
|
| 270 |
+
"max_position_embeddings": 64,
|
| 271 |
+
"model_type": "siglip_text_model",
|
| 272 |
+
"num_attention_heads": 12,
|
| 273 |
+
"num_hidden_layers": 12,
|
| 274 |
+
"projection_size": 768,
|
| 275 |
+
"torch_dtype": "float32",
|
| 276 |
+
"vocab_size": 256000
|
| 277 |
+
},
|
| 278 |
+
"torch_dtype": "float32",
|
| 279 |
+
"transformers_version": "4.51.0.dev0",
|
| 280 |
+
"vision_config": {
|
| 281 |
+
"attention_dropout": 0.0,
|
| 282 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 283 |
+
"hidden_size": 768,
|
| 284 |
+
"image_size": 224,
|
| 285 |
+
"intermediate_size": 3072,
|
| 286 |
+
"layer_norm_eps": 1e-06,
|
| 287 |
+
"model_type": "siglip_vision_model",
|
| 288 |
+
"num_attention_heads": 12,
|
| 289 |
+
"num_channels": 3,
|
| 290 |
+
"num_hidden_layers": 12,
|
| 291 |
+
"patch_size": 16,
|
| 292 |
+
"torch_dtype": "float32"
|
| 293 |
+
}
|
| 294 |
+
}
|
checkpoint-2320/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3c48077d6d68a4f4556e6e0343e3ad578bbcbdcf027767c3622cb6366f64e0c
|
| 3 |
+
size 371949432
|
checkpoint-2320/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d83496e9ccb3098762e648a386f65a2e2760abca9e2f2f4b923f1c1604508d8
|
| 3 |
+
size 687318522
|
checkpoint-2320/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-2320/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0d758056a18a79b432867affb13d3eadeec42e0600f59bc96c8f6bc6e0791e6
|
| 3 |
+
size 14244
|
checkpoint-2320/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9d23aced1d0188c1f587b6bf39dca60801fc311b2e80f1d9dfd3e72ebb2e1bf
|
| 3 |
+
size 1064
|
checkpoint-2320/trainer_state.json
ADDED
|
@@ -0,0 +1,112 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 2320,
|
| 3 |
+
"best_metric": 0.6425395607948303,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-2320",
|
| 5 |
+
"epoch": 5.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 2320,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 1.0,
|
| 14 |
+
"eval_accuracy": 0.6359832635983264,
|
| 15 |
+
"eval_loss": 2.0701308250427246,
|
| 16 |
+
"eval_model_preparation_time": 0.0031,
|
| 17 |
+
"eval_runtime": 215.6864,
|
| 18 |
+
"eval_samples_per_second": 68.702,
|
| 19 |
+
"eval_steps_per_second": 8.591,
|
| 20 |
+
"step": 464
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 1.0775862068965518,
|
| 24 |
+
"grad_norm": 49.15087890625,
|
| 25 |
+
"learning_rate": 2.4052863436123347e-06,
|
| 26 |
+
"loss": 3.5107,
|
| 27 |
+
"step": 500
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 2.0,
|
| 31 |
+
"eval_accuracy": 0.8090160615467674,
|
| 32 |
+
"eval_loss": 1.0770337581634521,
|
| 33 |
+
"eval_model_preparation_time": 0.0031,
|
| 34 |
+
"eval_runtime": 213.0035,
|
| 35 |
+
"eval_samples_per_second": 69.567,
|
| 36 |
+
"eval_steps_per_second": 8.699,
|
| 37 |
+
"step": 928
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 2.1551724137931036,
|
| 41 |
+
"grad_norm": 57.83103561401367,
|
| 42 |
+
"learning_rate": 1.7444933920704845e-06,
|
| 43 |
+
"loss": 1.3776,
|
| 44 |
+
"step": 1000
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 3.0,
|
| 48 |
+
"eval_accuracy": 0.8484275880685652,
|
| 49 |
+
"eval_loss": 0.7880541086196899,
|
| 50 |
+
"eval_model_preparation_time": 0.0031,
|
| 51 |
+
"eval_runtime": 213.1605,
|
| 52 |
+
"eval_samples_per_second": 69.516,
|
| 53 |
+
"eval_steps_per_second": 8.693,
|
| 54 |
+
"step": 1392
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 3.2327586206896552,
|
| 58 |
+
"grad_norm": 65.75613403320312,
|
| 59 |
+
"learning_rate": 1.0837004405286343e-06,
|
| 60 |
+
"loss": 0.8854,
|
| 61 |
+
"step": 1500
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"epoch": 4.0,
|
| 65 |
+
"eval_accuracy": 0.8636118234579565,
|
| 66 |
+
"eval_loss": 0.6761099696159363,
|
| 67 |
+
"eval_model_preparation_time": 0.0031,
|
| 68 |
+
"eval_runtime": 212.8345,
|
| 69 |
+
"eval_samples_per_second": 69.622,
|
| 70 |
+
"eval_steps_per_second": 8.706,
|
| 71 |
+
"step": 1856
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 4.310344827586207,
|
| 75 |
+
"grad_norm": 47.517181396484375,
|
| 76 |
+
"learning_rate": 4.229074889867842e-07,
|
| 77 |
+
"loss": 0.7222,
|
| 78 |
+
"step": 2000
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"epoch": 5.0,
|
| 82 |
+
"eval_accuracy": 0.869145633688757,
|
| 83 |
+
"eval_loss": 0.6425395607948303,
|
| 84 |
+
"eval_model_preparation_time": 0.0031,
|
| 85 |
+
"eval_runtime": 216.0108,
|
| 86 |
+
"eval_samples_per_second": 68.598,
|
| 87 |
+
"eval_steps_per_second": 8.578,
|
| 88 |
+
"step": 2320
|
| 89 |
+
}
|
| 90 |
+
],
|
| 91 |
+
"logging_steps": 500,
|
| 92 |
+
"max_steps": 2320,
|
| 93 |
+
"num_input_tokens_seen": 0,
|
| 94 |
+
"num_train_epochs": 5,
|
| 95 |
+
"save_steps": 500,
|
| 96 |
+
"stateful_callbacks": {
|
| 97 |
+
"TrainerControl": {
|
| 98 |
+
"args": {
|
| 99 |
+
"should_epoch_stop": false,
|
| 100 |
+
"should_evaluate": false,
|
| 101 |
+
"should_log": false,
|
| 102 |
+
"should_save": true,
|
| 103 |
+
"should_training_stop": true
|
| 104 |
+
},
|
| 105 |
+
"attributes": {}
|
| 106 |
+
}
|
| 107 |
+
},
|
| 108 |
+
"total_flos": 6.211825500522701e+18,
|
| 109 |
+
"train_batch_size": 32,
|
| 110 |
+
"trial_name": null,
|
| 111 |
+
"trial_params": null
|
| 112 |
+
}
|
checkpoint-2320/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d60dfc3a7f3feaaf5fab64c05e89c94efd07d977dc0d1cf7a28f5e86e8cbf7e5
|
| 3 |
+
size 5304
|
checkpoint-464/config.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "aircraft_carrier",
|
| 7 |
+
"1": "alarm_clock",
|
| 8 |
+
"2": "ant",
|
| 9 |
+
"3": "anvil",
|
| 10 |
+
"4": "asparagus",
|
| 11 |
+
"5": "axe",
|
| 12 |
+
"6": "banana",
|
| 13 |
+
"7": "basket",
|
| 14 |
+
"8": "bathtub",
|
| 15 |
+
"9": "bear",
|
| 16 |
+
"10": "bee",
|
| 17 |
+
"11": "bird",
|
| 18 |
+
"12": "blackberry",
|
| 19 |
+
"13": "blueberry",
|
| 20 |
+
"14": "bottlecap",
|
| 21 |
+
"15": "broccoli",
|
| 22 |
+
"16": "bus",
|
| 23 |
+
"17": "butterfly",
|
| 24 |
+
"18": "cactus",
|
| 25 |
+
"19": "cake",
|
| 26 |
+
"20": "calculator",
|
| 27 |
+
"21": "camel",
|
| 28 |
+
"22": "camera",
|
| 29 |
+
"23": "candle",
|
| 30 |
+
"24": "cannon",
|
| 31 |
+
"25": "canoe",
|
| 32 |
+
"26": "carrot",
|
| 33 |
+
"27": "castle",
|
| 34 |
+
"28": "cat",
|
| 35 |
+
"29": "ceiling_fan",
|
| 36 |
+
"30": "cell_phone",
|
| 37 |
+
"31": "cello",
|
| 38 |
+
"32": "chair",
|
| 39 |
+
"33": "chandelier",
|
| 40 |
+
"34": "coffee_cup",
|
| 41 |
+
"35": "compass",
|
| 42 |
+
"36": "computer",
|
| 43 |
+
"37": "cow",
|
| 44 |
+
"38": "crab",
|
| 45 |
+
"39": "crocodile",
|
| 46 |
+
"40": "cruise_ship",
|
| 47 |
+
"41": "dog",
|
| 48 |
+
"42": "dolphin",
|
| 49 |
+
"43": "dragon",
|
| 50 |
+
"44": "drums",
|
| 51 |
+
"45": "duck",
|
| 52 |
+
"46": "dumbbell",
|
| 53 |
+
"47": "elephant",
|
| 54 |
+
"48": "eyeglasses",
|
| 55 |
+
"49": "feather",
|
| 56 |
+
"50": "fence",
|
| 57 |
+
"51": "fish",
|
| 58 |
+
"52": "flamingo",
|
| 59 |
+
"53": "flower",
|
| 60 |
+
"54": "foot",
|
| 61 |
+
"55": "fork",
|
| 62 |
+
"56": "frog",
|
| 63 |
+
"57": "giraffe",
|
| 64 |
+
"58": "goatee",
|
| 65 |
+
"59": "grapes",
|
| 66 |
+
"60": "guitar",
|
| 67 |
+
"61": "hammer",
|
| 68 |
+
"62": "helicopter",
|
| 69 |
+
"63": "helmet",
|
| 70 |
+
"64": "horse",
|
| 71 |
+
"65": "kangaroo",
|
| 72 |
+
"66": "lantern",
|
| 73 |
+
"67": "laptop",
|
| 74 |
+
"68": "leaf",
|
| 75 |
+
"69": "lion",
|
| 76 |
+
"70": "lipstick",
|
| 77 |
+
"71": "lobster",
|
| 78 |
+
"72": "microphone",
|
| 79 |
+
"73": "monkey",
|
| 80 |
+
"74": "mosquito",
|
| 81 |
+
"75": "mouse",
|
| 82 |
+
"76": "mug",
|
| 83 |
+
"77": "mushroom",
|
| 84 |
+
"78": "onion",
|
| 85 |
+
"79": "panda",
|
| 86 |
+
"80": "peanut",
|
| 87 |
+
"81": "pear",
|
| 88 |
+
"82": "peas",
|
| 89 |
+
"83": "pencil",
|
| 90 |
+
"84": "penguin",
|
| 91 |
+
"85": "pig",
|
| 92 |
+
"86": "pillow",
|
| 93 |
+
"87": "pineapple",
|
| 94 |
+
"88": "potato",
|
| 95 |
+
"89": "power_outlet",
|
| 96 |
+
"90": "purse",
|
| 97 |
+
"91": "rabbit",
|
| 98 |
+
"92": "raccoon",
|
| 99 |
+
"93": "rhinoceros",
|
| 100 |
+
"94": "rifle",
|
| 101 |
+
"95": "saxophone",
|
| 102 |
+
"96": "screwdriver",
|
| 103 |
+
"97": "sea_turtle",
|
| 104 |
+
"98": "see_saw",
|
| 105 |
+
"99": "sheep",
|
| 106 |
+
"100": "shoe",
|
| 107 |
+
"101": "skateboard",
|
| 108 |
+
"102": "snake",
|
| 109 |
+
"103": "speedboat",
|
| 110 |
+
"104": "spider",
|
| 111 |
+
"105": "squirrel",
|
| 112 |
+
"106": "strawberry",
|
| 113 |
+
"107": "streetlight",
|
| 114 |
+
"108": "string_bean",
|
| 115 |
+
"109": "submarine",
|
| 116 |
+
"110": "swan",
|
| 117 |
+
"111": "table",
|
| 118 |
+
"112": "teapot",
|
| 119 |
+
"113": "teddy-bear",
|
| 120 |
+
"114": "television",
|
| 121 |
+
"115": "the_Eiffel_Tower",
|
| 122 |
+
"116": "the_Great_Wall_of_China",
|
| 123 |
+
"117": "tiger",
|
| 124 |
+
"118": "toe",
|
| 125 |
+
"119": "train",
|
| 126 |
+
"120": "truck",
|
| 127 |
+
"121": "umbrella",
|
| 128 |
+
"122": "vase",
|
| 129 |
+
"123": "watermelon",
|
| 130 |
+
"124": "whale",
|
| 131 |
+
"125": "zebra"
|
| 132 |
+
},
|
| 133 |
+
"initializer_factor": 1.0,
|
| 134 |
+
"label2id": {
|
| 135 |
+
"aircraft_carrier": 0,
|
| 136 |
+
"alarm_clock": 1,
|
| 137 |
+
"ant": 2,
|
| 138 |
+
"anvil": 3,
|
| 139 |
+
"asparagus": 4,
|
| 140 |
+
"axe": 5,
|
| 141 |
+
"banana": 6,
|
| 142 |
+
"basket": 7,
|
| 143 |
+
"bathtub": 8,
|
| 144 |
+
"bear": 9,
|
| 145 |
+
"bee": 10,
|
| 146 |
+
"bird": 11,
|
| 147 |
+
"blackberry": 12,
|
| 148 |
+
"blueberry": 13,
|
| 149 |
+
"bottlecap": 14,
|
| 150 |
+
"broccoli": 15,
|
| 151 |
+
"bus": 16,
|
| 152 |
+
"butterfly": 17,
|
| 153 |
+
"cactus": 18,
|
| 154 |
+
"cake": 19,
|
| 155 |
+
"calculator": 20,
|
| 156 |
+
"camel": 21,
|
| 157 |
+
"camera": 22,
|
| 158 |
+
"candle": 23,
|
| 159 |
+
"cannon": 24,
|
| 160 |
+
"canoe": 25,
|
| 161 |
+
"carrot": 26,
|
| 162 |
+
"castle": 27,
|
| 163 |
+
"cat": 28,
|
| 164 |
+
"ceiling_fan": 29,
|
| 165 |
+
"cell_phone": 30,
|
| 166 |
+
"cello": 31,
|
| 167 |
+
"chair": 32,
|
| 168 |
+
"chandelier": 33,
|
| 169 |
+
"coffee_cup": 34,
|
| 170 |
+
"compass": 35,
|
| 171 |
+
"computer": 36,
|
| 172 |
+
"cow": 37,
|
| 173 |
+
"crab": 38,
|
| 174 |
+
"crocodile": 39,
|
| 175 |
+
"cruise_ship": 40,
|
| 176 |
+
"dog": 41,
|
| 177 |
+
"dolphin": 42,
|
| 178 |
+
"dragon": 43,
|
| 179 |
+
"drums": 44,
|
| 180 |
+
"duck": 45,
|
| 181 |
+
"dumbbell": 46,
|
| 182 |
+
"elephant": 47,
|
| 183 |
+
"eyeglasses": 48,
|
| 184 |
+
"feather": 49,
|
| 185 |
+
"fence": 50,
|
| 186 |
+
"fish": 51,
|
| 187 |
+
"flamingo": 52,
|
| 188 |
+
"flower": 53,
|
| 189 |
+
"foot": 54,
|
| 190 |
+
"fork": 55,
|
| 191 |
+
"frog": 56,
|
| 192 |
+
"giraffe": 57,
|
| 193 |
+
"goatee": 58,
|
| 194 |
+
"grapes": 59,
|
| 195 |
+
"guitar": 60,
|
| 196 |
+
"hammer": 61,
|
| 197 |
+
"helicopter": 62,
|
| 198 |
+
"helmet": 63,
|
| 199 |
+
"horse": 64,
|
| 200 |
+
"kangaroo": 65,
|
| 201 |
+
"lantern": 66,
|
| 202 |
+
"laptop": 67,
|
| 203 |
+
"leaf": 68,
|
| 204 |
+
"lion": 69,
|
| 205 |
+
"lipstick": 70,
|
| 206 |
+
"lobster": 71,
|
| 207 |
+
"microphone": 72,
|
| 208 |
+
"monkey": 73,
|
| 209 |
+
"mosquito": 74,
|
| 210 |
+
"mouse": 75,
|
| 211 |
+
"mug": 76,
|
| 212 |
+
"mushroom": 77,
|
| 213 |
+
"onion": 78,
|
| 214 |
+
"panda": 79,
|
| 215 |
+
"peanut": 80,
|
| 216 |
+
"pear": 81,
|
| 217 |
+
"peas": 82,
|
| 218 |
+
"pencil": 83,
|
| 219 |
+
"penguin": 84,
|
| 220 |
+
"pig": 85,
|
| 221 |
+
"pillow": 86,
|
| 222 |
+
"pineapple": 87,
|
| 223 |
+
"potato": 88,
|
| 224 |
+
"power_outlet": 89,
|
| 225 |
+
"purse": 90,
|
| 226 |
+
"rabbit": 91,
|
| 227 |
+
"raccoon": 92,
|
| 228 |
+
"rhinoceros": 93,
|
| 229 |
+
"rifle": 94,
|
| 230 |
+
"saxophone": 95,
|
| 231 |
+
"screwdriver": 96,
|
| 232 |
+
"sea_turtle": 97,
|
| 233 |
+
"see_saw": 98,
|
| 234 |
+
"sheep": 99,
|
| 235 |
+
"shoe": 100,
|
| 236 |
+
"skateboard": 101,
|
| 237 |
+
"snake": 102,
|
| 238 |
+
"speedboat": 103,
|
| 239 |
+
"spider": 104,
|
| 240 |
+
"squirrel": 105,
|
| 241 |
+
"strawberry": 106,
|
| 242 |
+
"streetlight": 107,
|
| 243 |
+
"string_bean": 108,
|
| 244 |
+
"submarine": 109,
|
| 245 |
+
"swan": 110,
|
| 246 |
+
"table": 111,
|
| 247 |
+
"teapot": 112,
|
| 248 |
+
"teddy-bear": 113,
|
| 249 |
+
"television": 114,
|
| 250 |
+
"the_Eiffel_Tower": 115,
|
| 251 |
+
"the_Great_Wall_of_China": 116,
|
| 252 |
+
"tiger": 117,
|
| 253 |
+
"toe": 118,
|
| 254 |
+
"train": 119,
|
| 255 |
+
"truck": 120,
|
| 256 |
+
"umbrella": 121,
|
| 257 |
+
"vase": 122,
|
| 258 |
+
"watermelon": 123,
|
| 259 |
+
"whale": 124,
|
| 260 |
+
"zebra": 125
|
| 261 |
+
},
|
| 262 |
+
"model_type": "siglip",
|
| 263 |
+
"problem_type": "single_label_classification",
|
| 264 |
+
"text_config": {
|
| 265 |
+
"attention_dropout": 0.0,
|
| 266 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 267 |
+
"hidden_size": 768,
|
| 268 |
+
"intermediate_size": 3072,
|
| 269 |
+
"layer_norm_eps": 1e-06,
|
| 270 |
+
"max_position_embeddings": 64,
|
| 271 |
+
"model_type": "siglip_text_model",
|
| 272 |
+
"num_attention_heads": 12,
|
| 273 |
+
"num_hidden_layers": 12,
|
| 274 |
+
"projection_size": 768,
|
| 275 |
+
"torch_dtype": "float32",
|
| 276 |
+
"vocab_size": 256000
|
| 277 |
+
},
|
| 278 |
+
"torch_dtype": "float32",
|
| 279 |
+
"transformers_version": "4.51.0.dev0",
|
| 280 |
+
"vision_config": {
|
| 281 |
+
"attention_dropout": 0.0,
|
| 282 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 283 |
+
"hidden_size": 768,
|
| 284 |
+
"image_size": 224,
|
| 285 |
+
"intermediate_size": 3072,
|
| 286 |
+
"layer_norm_eps": 1e-06,
|
| 287 |
+
"model_type": "siglip_vision_model",
|
| 288 |
+
"num_attention_heads": 12,
|
| 289 |
+
"num_channels": 3,
|
| 290 |
+
"num_hidden_layers": 12,
|
| 291 |
+
"patch_size": 16,
|
| 292 |
+
"torch_dtype": "float32"
|
| 293 |
+
}
|
| 294 |
+
}
|
checkpoint-464/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5de2b6b9588b3364539be635283c74a5ce5c0a0498b6ebd72df451826b204119
|
| 3 |
+
size 371949432
|
checkpoint-464/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2162ee4999aef1f7c9cabf7addf7d4fb6e8fb6aa7c57c025e3c158954e1cf531
|
| 3 |
+
size 687318522
|
checkpoint-464/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-464/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d4c8bdc9624035b7c58983b3149870e44adcc08330e275709e5a7b52b9fc831
|
| 3 |
+
size 14244
|
checkpoint-464/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01dba65e381eb16cbaad2e89bc489f7994479ad04028b9d06367b9af61c412bb
|
| 3 |
+
size 1064
|
checkpoint-464/trainer_state.json
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 464,
|
| 3 |
+
"best_metric": 2.0701308250427246,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-464",
|
| 5 |
+
"epoch": 1.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 464,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 1.0,
|
| 14 |
+
"eval_accuracy": 0.6359832635983264,
|
| 15 |
+
"eval_loss": 2.0701308250427246,
|
| 16 |
+
"eval_model_preparation_time": 0.0031,
|
| 17 |
+
"eval_runtime": 215.6864,
|
| 18 |
+
"eval_samples_per_second": 68.702,
|
| 19 |
+
"eval_steps_per_second": 8.591,
|
| 20 |
+
"step": 464
|
| 21 |
+
}
|
| 22 |
+
],
|
| 23 |
+
"logging_steps": 500,
|
| 24 |
+
"max_steps": 2320,
|
| 25 |
+
"num_input_tokens_seen": 0,
|
| 26 |
+
"num_train_epochs": 5,
|
| 27 |
+
"save_steps": 500,
|
| 28 |
+
"stateful_callbacks": {
|
| 29 |
+
"TrainerControl": {
|
| 30 |
+
"args": {
|
| 31 |
+
"should_epoch_stop": false,
|
| 32 |
+
"should_evaluate": false,
|
| 33 |
+
"should_log": false,
|
| 34 |
+
"should_save": true,
|
| 35 |
+
"should_training_stop": false
|
| 36 |
+
},
|
| 37 |
+
"attributes": {}
|
| 38 |
+
}
|
| 39 |
+
},
|
| 40 |
+
"total_flos": 1.2423651001045402e+18,
|
| 41 |
+
"train_batch_size": 32,
|
| 42 |
+
"trial_name": null,
|
| 43 |
+
"trial_params": null
|
| 44 |
+
}
|
checkpoint-464/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d60dfc3a7f3feaaf5fab64c05e89c94efd07d977dc0d1cf7a28f5e86e8cbf7e5
|
| 3 |
+
size 5304
|
checkpoint-928/config.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "aircraft_carrier",
|
| 7 |
+
"1": "alarm_clock",
|
| 8 |
+
"2": "ant",
|
| 9 |
+
"3": "anvil",
|
| 10 |
+
"4": "asparagus",
|
| 11 |
+
"5": "axe",
|
| 12 |
+
"6": "banana",
|
| 13 |
+
"7": "basket",
|
| 14 |
+
"8": "bathtub",
|
| 15 |
+
"9": "bear",
|
| 16 |
+
"10": "bee",
|
| 17 |
+
"11": "bird",
|
| 18 |
+
"12": "blackberry",
|
| 19 |
+
"13": "blueberry",
|
| 20 |
+
"14": "bottlecap",
|
| 21 |
+
"15": "broccoli",
|
| 22 |
+
"16": "bus",
|
| 23 |
+
"17": "butterfly",
|
| 24 |
+
"18": "cactus",
|
| 25 |
+
"19": "cake",
|
| 26 |
+
"20": "calculator",
|
| 27 |
+
"21": "camel",
|
| 28 |
+
"22": "camera",
|
| 29 |
+
"23": "candle",
|
| 30 |
+
"24": "cannon",
|
| 31 |
+
"25": "canoe",
|
| 32 |
+
"26": "carrot",
|
| 33 |
+
"27": "castle",
|
| 34 |
+
"28": "cat",
|
| 35 |
+
"29": "ceiling_fan",
|
| 36 |
+
"30": "cell_phone",
|
| 37 |
+
"31": "cello",
|
| 38 |
+
"32": "chair",
|
| 39 |
+
"33": "chandelier",
|
| 40 |
+
"34": "coffee_cup",
|
| 41 |
+
"35": "compass",
|
| 42 |
+
"36": "computer",
|
| 43 |
+
"37": "cow",
|
| 44 |
+
"38": "crab",
|
| 45 |
+
"39": "crocodile",
|
| 46 |
+
"40": "cruise_ship",
|
| 47 |
+
"41": "dog",
|
| 48 |
+
"42": "dolphin",
|
| 49 |
+
"43": "dragon",
|
| 50 |
+
"44": "drums",
|
| 51 |
+
"45": "duck",
|
| 52 |
+
"46": "dumbbell",
|
| 53 |
+
"47": "elephant",
|
| 54 |
+
"48": "eyeglasses",
|
| 55 |
+
"49": "feather",
|
| 56 |
+
"50": "fence",
|
| 57 |
+
"51": "fish",
|
| 58 |
+
"52": "flamingo",
|
| 59 |
+
"53": "flower",
|
| 60 |
+
"54": "foot",
|
| 61 |
+
"55": "fork",
|
| 62 |
+
"56": "frog",
|
| 63 |
+
"57": "giraffe",
|
| 64 |
+
"58": "goatee",
|
| 65 |
+
"59": "grapes",
|
| 66 |
+
"60": "guitar",
|
| 67 |
+
"61": "hammer",
|
| 68 |
+
"62": "helicopter",
|
| 69 |
+
"63": "helmet",
|
| 70 |
+
"64": "horse",
|
| 71 |
+
"65": "kangaroo",
|
| 72 |
+
"66": "lantern",
|
| 73 |
+
"67": "laptop",
|
| 74 |
+
"68": "leaf",
|
| 75 |
+
"69": "lion",
|
| 76 |
+
"70": "lipstick",
|
| 77 |
+
"71": "lobster",
|
| 78 |
+
"72": "microphone",
|
| 79 |
+
"73": "monkey",
|
| 80 |
+
"74": "mosquito",
|
| 81 |
+
"75": "mouse",
|
| 82 |
+
"76": "mug",
|
| 83 |
+
"77": "mushroom",
|
| 84 |
+
"78": "onion",
|
| 85 |
+
"79": "panda",
|
| 86 |
+
"80": "peanut",
|
| 87 |
+
"81": "pear",
|
| 88 |
+
"82": "peas",
|
| 89 |
+
"83": "pencil",
|
| 90 |
+
"84": "penguin",
|
| 91 |
+
"85": "pig",
|
| 92 |
+
"86": "pillow",
|
| 93 |
+
"87": "pineapple",
|
| 94 |
+
"88": "potato",
|
| 95 |
+
"89": "power_outlet",
|
| 96 |
+
"90": "purse",
|
| 97 |
+
"91": "rabbit",
|
| 98 |
+
"92": "raccoon",
|
| 99 |
+
"93": "rhinoceros",
|
| 100 |
+
"94": "rifle",
|
| 101 |
+
"95": "saxophone",
|
| 102 |
+
"96": "screwdriver",
|
| 103 |
+
"97": "sea_turtle",
|
| 104 |
+
"98": "see_saw",
|
| 105 |
+
"99": "sheep",
|
| 106 |
+
"100": "shoe",
|
| 107 |
+
"101": "skateboard",
|
| 108 |
+
"102": "snake",
|
| 109 |
+
"103": "speedboat",
|
| 110 |
+
"104": "spider",
|
| 111 |
+
"105": "squirrel",
|
| 112 |
+
"106": "strawberry",
|
| 113 |
+
"107": "streetlight",
|
| 114 |
+
"108": "string_bean",
|
| 115 |
+
"109": "submarine",
|
| 116 |
+
"110": "swan",
|
| 117 |
+
"111": "table",
|
| 118 |
+
"112": "teapot",
|
| 119 |
+
"113": "teddy-bear",
|
| 120 |
+
"114": "television",
|
| 121 |
+
"115": "the_Eiffel_Tower",
|
| 122 |
+
"116": "the_Great_Wall_of_China",
|
| 123 |
+
"117": "tiger",
|
| 124 |
+
"118": "toe",
|
| 125 |
+
"119": "train",
|
| 126 |
+
"120": "truck",
|
| 127 |
+
"121": "umbrella",
|
| 128 |
+
"122": "vase",
|
| 129 |
+
"123": "watermelon",
|
| 130 |
+
"124": "whale",
|
| 131 |
+
"125": "zebra"
|
| 132 |
+
},
|
| 133 |
+
"initializer_factor": 1.0,
|
| 134 |
+
"label2id": {
|
| 135 |
+
"aircraft_carrier": 0,
|
| 136 |
+
"alarm_clock": 1,
|
| 137 |
+
"ant": 2,
|
| 138 |
+
"anvil": 3,
|
| 139 |
+
"asparagus": 4,
|
| 140 |
+
"axe": 5,
|
| 141 |
+
"banana": 6,
|
| 142 |
+
"basket": 7,
|
| 143 |
+
"bathtub": 8,
|
| 144 |
+
"bear": 9,
|
| 145 |
+
"bee": 10,
|
| 146 |
+
"bird": 11,
|
| 147 |
+
"blackberry": 12,
|
| 148 |
+
"blueberry": 13,
|
| 149 |
+
"bottlecap": 14,
|
| 150 |
+
"broccoli": 15,
|
| 151 |
+
"bus": 16,
|
| 152 |
+
"butterfly": 17,
|
| 153 |
+
"cactus": 18,
|
| 154 |
+
"cake": 19,
|
| 155 |
+
"calculator": 20,
|
| 156 |
+
"camel": 21,
|
| 157 |
+
"camera": 22,
|
| 158 |
+
"candle": 23,
|
| 159 |
+
"cannon": 24,
|
| 160 |
+
"canoe": 25,
|
| 161 |
+
"carrot": 26,
|
| 162 |
+
"castle": 27,
|
| 163 |
+
"cat": 28,
|
| 164 |
+
"ceiling_fan": 29,
|
| 165 |
+
"cell_phone": 30,
|
| 166 |
+
"cello": 31,
|
| 167 |
+
"chair": 32,
|
| 168 |
+
"chandelier": 33,
|
| 169 |
+
"coffee_cup": 34,
|
| 170 |
+
"compass": 35,
|
| 171 |
+
"computer": 36,
|
| 172 |
+
"cow": 37,
|
| 173 |
+
"crab": 38,
|
| 174 |
+
"crocodile": 39,
|
| 175 |
+
"cruise_ship": 40,
|
| 176 |
+
"dog": 41,
|
| 177 |
+
"dolphin": 42,
|
| 178 |
+
"dragon": 43,
|
| 179 |
+
"drums": 44,
|
| 180 |
+
"duck": 45,
|
| 181 |
+
"dumbbell": 46,
|
| 182 |
+
"elephant": 47,
|
| 183 |
+
"eyeglasses": 48,
|
| 184 |
+
"feather": 49,
|
| 185 |
+
"fence": 50,
|
| 186 |
+
"fish": 51,
|
| 187 |
+
"flamingo": 52,
|
| 188 |
+
"flower": 53,
|
| 189 |
+
"foot": 54,
|
| 190 |
+
"fork": 55,
|
| 191 |
+
"frog": 56,
|
| 192 |
+
"giraffe": 57,
|
| 193 |
+
"goatee": 58,
|
| 194 |
+
"grapes": 59,
|
| 195 |
+
"guitar": 60,
|
| 196 |
+
"hammer": 61,
|
| 197 |
+
"helicopter": 62,
|
| 198 |
+
"helmet": 63,
|
| 199 |
+
"horse": 64,
|
| 200 |
+
"kangaroo": 65,
|
| 201 |
+
"lantern": 66,
|
| 202 |
+
"laptop": 67,
|
| 203 |
+
"leaf": 68,
|
| 204 |
+
"lion": 69,
|
| 205 |
+
"lipstick": 70,
|
| 206 |
+
"lobster": 71,
|
| 207 |
+
"microphone": 72,
|
| 208 |
+
"monkey": 73,
|
| 209 |
+
"mosquito": 74,
|
| 210 |
+
"mouse": 75,
|
| 211 |
+
"mug": 76,
|
| 212 |
+
"mushroom": 77,
|
| 213 |
+
"onion": 78,
|
| 214 |
+
"panda": 79,
|
| 215 |
+
"peanut": 80,
|
| 216 |
+
"pear": 81,
|
| 217 |
+
"peas": 82,
|
| 218 |
+
"pencil": 83,
|
| 219 |
+
"penguin": 84,
|
| 220 |
+
"pig": 85,
|
| 221 |
+
"pillow": 86,
|
| 222 |
+
"pineapple": 87,
|
| 223 |
+
"potato": 88,
|
| 224 |
+
"power_outlet": 89,
|
| 225 |
+
"purse": 90,
|
| 226 |
+
"rabbit": 91,
|
| 227 |
+
"raccoon": 92,
|
| 228 |
+
"rhinoceros": 93,
|
| 229 |
+
"rifle": 94,
|
| 230 |
+
"saxophone": 95,
|
| 231 |
+
"screwdriver": 96,
|
| 232 |
+
"sea_turtle": 97,
|
| 233 |
+
"see_saw": 98,
|
| 234 |
+
"sheep": 99,
|
| 235 |
+
"shoe": 100,
|
| 236 |
+
"skateboard": 101,
|
| 237 |
+
"snake": 102,
|
| 238 |
+
"speedboat": 103,
|
| 239 |
+
"spider": 104,
|
| 240 |
+
"squirrel": 105,
|
| 241 |
+
"strawberry": 106,
|
| 242 |
+
"streetlight": 107,
|
| 243 |
+
"string_bean": 108,
|
| 244 |
+
"submarine": 109,
|
| 245 |
+
"swan": 110,
|
| 246 |
+
"table": 111,
|
| 247 |
+
"teapot": 112,
|
| 248 |
+
"teddy-bear": 113,
|
| 249 |
+
"television": 114,
|
| 250 |
+
"the_Eiffel_Tower": 115,
|
| 251 |
+
"the_Great_Wall_of_China": 116,
|
| 252 |
+
"tiger": 117,
|
| 253 |
+
"toe": 118,
|
| 254 |
+
"train": 119,
|
| 255 |
+
"truck": 120,
|
| 256 |
+
"umbrella": 121,
|
| 257 |
+
"vase": 122,
|
| 258 |
+
"watermelon": 123,
|
| 259 |
+
"whale": 124,
|
| 260 |
+
"zebra": 125
|
| 261 |
+
},
|
| 262 |
+
"model_type": "siglip",
|
| 263 |
+
"problem_type": "single_label_classification",
|
| 264 |
+
"text_config": {
|
| 265 |
+
"attention_dropout": 0.0,
|
| 266 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 267 |
+
"hidden_size": 768,
|
| 268 |
+
"intermediate_size": 3072,
|
| 269 |
+
"layer_norm_eps": 1e-06,
|
| 270 |
+
"max_position_embeddings": 64,
|
| 271 |
+
"model_type": "siglip_text_model",
|
| 272 |
+
"num_attention_heads": 12,
|
| 273 |
+
"num_hidden_layers": 12,
|
| 274 |
+
"projection_size": 768,
|
| 275 |
+
"torch_dtype": "float32",
|
| 276 |
+
"vocab_size": 256000
|
| 277 |
+
},
|
| 278 |
+
"torch_dtype": "float32",
|
| 279 |
+
"transformers_version": "4.51.0.dev0",
|
| 280 |
+
"vision_config": {
|
| 281 |
+
"attention_dropout": 0.0,
|
| 282 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 283 |
+
"hidden_size": 768,
|
| 284 |
+
"image_size": 224,
|
| 285 |
+
"intermediate_size": 3072,
|
| 286 |
+
"layer_norm_eps": 1e-06,
|
| 287 |
+
"model_type": "siglip_vision_model",
|
| 288 |
+
"num_attention_heads": 12,
|
| 289 |
+
"num_channels": 3,
|
| 290 |
+
"num_hidden_layers": 12,
|
| 291 |
+
"patch_size": 16,
|
| 292 |
+
"torch_dtype": "float32"
|
| 293 |
+
}
|
| 294 |
+
}
|
checkpoint-928/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3758748686085d5795b8a20aa0266f8d7a8272402399ea9deb593434ee44ce73
|
| 3 |
+
size 371949432
|
checkpoint-928/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a67928d7c2414fc14ae4ed8e823fc4cedfedd94a487c63238134dfd90e268ec
|
| 3 |
+
size 687318522
|
checkpoint-928/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-928/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d2d10a1eabe4246fdd8425397fcbec8897bd0ab9d578470695da8deeaf4719b4
|
| 3 |
+
size 14244
|
checkpoint-928/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6877dbe8702a5e6a6128648d9dd20ff0f557202531aa053af9e2ac2cd66b0cb1
|
| 3 |
+
size 1064
|
checkpoint-928/trainer_state.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 928,
|
| 3 |
+
"best_metric": 1.0770337581634521,
|
| 4 |
+
"best_model_checkpoint": "siglip2-finetune-full/checkpoint-928",
|
| 5 |
+
"epoch": 2.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 928,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 1.0,
|
| 14 |
+
"eval_accuracy": 0.6359832635983264,
|
| 15 |
+
"eval_loss": 2.0701308250427246,
|
| 16 |
+
"eval_model_preparation_time": 0.0031,
|
| 17 |
+
"eval_runtime": 215.6864,
|
| 18 |
+
"eval_samples_per_second": 68.702,
|
| 19 |
+
"eval_steps_per_second": 8.591,
|
| 20 |
+
"step": 464
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 1.0775862068965518,
|
| 24 |
+
"grad_norm": 49.15087890625,
|
| 25 |
+
"learning_rate": 2.4052863436123347e-06,
|
| 26 |
+
"loss": 3.5107,
|
| 27 |
+
"step": 500
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 2.0,
|
| 31 |
+
"eval_accuracy": 0.8090160615467674,
|
| 32 |
+
"eval_loss": 1.0770337581634521,
|
| 33 |
+
"eval_model_preparation_time": 0.0031,
|
| 34 |
+
"eval_runtime": 213.0035,
|
| 35 |
+
"eval_samples_per_second": 69.567,
|
| 36 |
+
"eval_steps_per_second": 8.699,
|
| 37 |
+
"step": 928
|
| 38 |
+
}
|
| 39 |
+
],
|
| 40 |
+
"logging_steps": 500,
|
| 41 |
+
"max_steps": 2320,
|
| 42 |
+
"num_input_tokens_seen": 0,
|
| 43 |
+
"num_train_epochs": 5,
|
| 44 |
+
"save_steps": 500,
|
| 45 |
+
"stateful_callbacks": {
|
| 46 |
+
"TrainerControl": {
|
| 47 |
+
"args": {
|
| 48 |
+
"should_epoch_stop": false,
|
| 49 |
+
"should_evaluate": false,
|
| 50 |
+
"should_log": false,
|
| 51 |
+
"should_save": true,
|
| 52 |
+
"should_training_stop": false
|
| 53 |
+
},
|
| 54 |
+
"attributes": {}
|
| 55 |
+
}
|
| 56 |
+
},
|
| 57 |
+
"total_flos": 2.4847302002090803e+18,
|
| 58 |
+
"train_batch_size": 32,
|
| 59 |
+
"trial_name": null,
|
| 60 |
+
"trial_params": null
|
| 61 |
+
}
|
checkpoint-928/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d60dfc3a7f3feaaf5fab64c05e89c94efd07d977dc0d1cf7a28f5e86e8cbf7e5
|
| 3 |
+
size 5304
|
config.json
ADDED
|
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SiglipForImageClassification"
|
| 4 |
+
],
|
| 5 |
+
"id2label": {
|
| 6 |
+
"0": "aircraft_carrier",
|
| 7 |
+
"1": "alarm_clock",
|
| 8 |
+
"2": "ant",
|
| 9 |
+
"3": "anvil",
|
| 10 |
+
"4": "asparagus",
|
| 11 |
+
"5": "axe",
|
| 12 |
+
"6": "banana",
|
| 13 |
+
"7": "basket",
|
| 14 |
+
"8": "bathtub",
|
| 15 |
+
"9": "bear",
|
| 16 |
+
"10": "bee",
|
| 17 |
+
"11": "bird",
|
| 18 |
+
"12": "blackberry",
|
| 19 |
+
"13": "blueberry",
|
| 20 |
+
"14": "bottlecap",
|
| 21 |
+
"15": "broccoli",
|
| 22 |
+
"16": "bus",
|
| 23 |
+
"17": "butterfly",
|
| 24 |
+
"18": "cactus",
|
| 25 |
+
"19": "cake",
|
| 26 |
+
"20": "calculator",
|
| 27 |
+
"21": "camel",
|
| 28 |
+
"22": "camera",
|
| 29 |
+
"23": "candle",
|
| 30 |
+
"24": "cannon",
|
| 31 |
+
"25": "canoe",
|
| 32 |
+
"26": "carrot",
|
| 33 |
+
"27": "castle",
|
| 34 |
+
"28": "cat",
|
| 35 |
+
"29": "ceiling_fan",
|
| 36 |
+
"30": "cell_phone",
|
| 37 |
+
"31": "cello",
|
| 38 |
+
"32": "chair",
|
| 39 |
+
"33": "chandelier",
|
| 40 |
+
"34": "coffee_cup",
|
| 41 |
+
"35": "compass",
|
| 42 |
+
"36": "computer",
|
| 43 |
+
"37": "cow",
|
| 44 |
+
"38": "crab",
|
| 45 |
+
"39": "crocodile",
|
| 46 |
+
"40": "cruise_ship",
|
| 47 |
+
"41": "dog",
|
| 48 |
+
"42": "dolphin",
|
| 49 |
+
"43": "dragon",
|
| 50 |
+
"44": "drums",
|
| 51 |
+
"45": "duck",
|
| 52 |
+
"46": "dumbbell",
|
| 53 |
+
"47": "elephant",
|
| 54 |
+
"48": "eyeglasses",
|
| 55 |
+
"49": "feather",
|
| 56 |
+
"50": "fence",
|
| 57 |
+
"51": "fish",
|
| 58 |
+
"52": "flamingo",
|
| 59 |
+
"53": "flower",
|
| 60 |
+
"54": "foot",
|
| 61 |
+
"55": "fork",
|
| 62 |
+
"56": "frog",
|
| 63 |
+
"57": "giraffe",
|
| 64 |
+
"58": "goatee",
|
| 65 |
+
"59": "grapes",
|
| 66 |
+
"60": "guitar",
|
| 67 |
+
"61": "hammer",
|
| 68 |
+
"62": "helicopter",
|
| 69 |
+
"63": "helmet",
|
| 70 |
+
"64": "horse",
|
| 71 |
+
"65": "kangaroo",
|
| 72 |
+
"66": "lantern",
|
| 73 |
+
"67": "laptop",
|
| 74 |
+
"68": "leaf",
|
| 75 |
+
"69": "lion",
|
| 76 |
+
"70": "lipstick",
|
| 77 |
+
"71": "lobster",
|
| 78 |
+
"72": "microphone",
|
| 79 |
+
"73": "monkey",
|
| 80 |
+
"74": "mosquito",
|
| 81 |
+
"75": "mouse",
|
| 82 |
+
"76": "mug",
|
| 83 |
+
"77": "mushroom",
|
| 84 |
+
"78": "onion",
|
| 85 |
+
"79": "panda",
|
| 86 |
+
"80": "peanut",
|
| 87 |
+
"81": "pear",
|
| 88 |
+
"82": "peas",
|
| 89 |
+
"83": "pencil",
|
| 90 |
+
"84": "penguin",
|
| 91 |
+
"85": "pig",
|
| 92 |
+
"86": "pillow",
|
| 93 |
+
"87": "pineapple",
|
| 94 |
+
"88": "potato",
|
| 95 |
+
"89": "power_outlet",
|
| 96 |
+
"90": "purse",
|
| 97 |
+
"91": "rabbit",
|
| 98 |
+
"92": "raccoon",
|
| 99 |
+
"93": "rhinoceros",
|
| 100 |
+
"94": "rifle",
|
| 101 |
+
"95": "saxophone",
|
| 102 |
+
"96": "screwdriver",
|
| 103 |
+
"97": "sea_turtle",
|
| 104 |
+
"98": "see_saw",
|
| 105 |
+
"99": "sheep",
|
| 106 |
+
"100": "shoe",
|
| 107 |
+
"101": "skateboard",
|
| 108 |
+
"102": "snake",
|
| 109 |
+
"103": "speedboat",
|
| 110 |
+
"104": "spider",
|
| 111 |
+
"105": "squirrel",
|
| 112 |
+
"106": "strawberry",
|
| 113 |
+
"107": "streetlight",
|
| 114 |
+
"108": "string_bean",
|
| 115 |
+
"109": "submarine",
|
| 116 |
+
"110": "swan",
|
| 117 |
+
"111": "table",
|
| 118 |
+
"112": "teapot",
|
| 119 |
+
"113": "teddy-bear",
|
| 120 |
+
"114": "television",
|
| 121 |
+
"115": "the_Eiffel_Tower",
|
| 122 |
+
"116": "the_Great_Wall_of_China",
|
| 123 |
+
"117": "tiger",
|
| 124 |
+
"118": "toe",
|
| 125 |
+
"119": "train",
|
| 126 |
+
"120": "truck",
|
| 127 |
+
"121": "umbrella",
|
| 128 |
+
"122": "vase",
|
| 129 |
+
"123": "watermelon",
|
| 130 |
+
"124": "whale",
|
| 131 |
+
"125": "zebra"
|
| 132 |
+
},
|
| 133 |
+
"initializer_factor": 1.0,
|
| 134 |
+
"label2id": {
|
| 135 |
+
"aircraft_carrier": 0,
|
| 136 |
+
"alarm_clock": 1,
|
| 137 |
+
"ant": 2,
|
| 138 |
+
"anvil": 3,
|
| 139 |
+
"asparagus": 4,
|
| 140 |
+
"axe": 5,
|
| 141 |
+
"banana": 6,
|
| 142 |
+
"basket": 7,
|
| 143 |
+
"bathtub": 8,
|
| 144 |
+
"bear": 9,
|
| 145 |
+
"bee": 10,
|
| 146 |
+
"bird": 11,
|
| 147 |
+
"blackberry": 12,
|
| 148 |
+
"blueberry": 13,
|
| 149 |
+
"bottlecap": 14,
|
| 150 |
+
"broccoli": 15,
|
| 151 |
+
"bus": 16,
|
| 152 |
+
"butterfly": 17,
|
| 153 |
+
"cactus": 18,
|
| 154 |
+
"cake": 19,
|
| 155 |
+
"calculator": 20,
|
| 156 |
+
"camel": 21,
|
| 157 |
+
"camera": 22,
|
| 158 |
+
"candle": 23,
|
| 159 |
+
"cannon": 24,
|
| 160 |
+
"canoe": 25,
|
| 161 |
+
"carrot": 26,
|
| 162 |
+
"castle": 27,
|
| 163 |
+
"cat": 28,
|
| 164 |
+
"ceiling_fan": 29,
|
| 165 |
+
"cell_phone": 30,
|
| 166 |
+
"cello": 31,
|
| 167 |
+
"chair": 32,
|
| 168 |
+
"chandelier": 33,
|
| 169 |
+
"coffee_cup": 34,
|
| 170 |
+
"compass": 35,
|
| 171 |
+
"computer": 36,
|
| 172 |
+
"cow": 37,
|
| 173 |
+
"crab": 38,
|
| 174 |
+
"crocodile": 39,
|
| 175 |
+
"cruise_ship": 40,
|
| 176 |
+
"dog": 41,
|
| 177 |
+
"dolphin": 42,
|
| 178 |
+
"dragon": 43,
|
| 179 |
+
"drums": 44,
|
| 180 |
+
"duck": 45,
|
| 181 |
+
"dumbbell": 46,
|
| 182 |
+
"elephant": 47,
|
| 183 |
+
"eyeglasses": 48,
|
| 184 |
+
"feather": 49,
|
| 185 |
+
"fence": 50,
|
| 186 |
+
"fish": 51,
|
| 187 |
+
"flamingo": 52,
|
| 188 |
+
"flower": 53,
|
| 189 |
+
"foot": 54,
|
| 190 |
+
"fork": 55,
|
| 191 |
+
"frog": 56,
|
| 192 |
+
"giraffe": 57,
|
| 193 |
+
"goatee": 58,
|
| 194 |
+
"grapes": 59,
|
| 195 |
+
"guitar": 60,
|
| 196 |
+
"hammer": 61,
|
| 197 |
+
"helicopter": 62,
|
| 198 |
+
"helmet": 63,
|
| 199 |
+
"horse": 64,
|
| 200 |
+
"kangaroo": 65,
|
| 201 |
+
"lantern": 66,
|
| 202 |
+
"laptop": 67,
|
| 203 |
+
"leaf": 68,
|
| 204 |
+
"lion": 69,
|
| 205 |
+
"lipstick": 70,
|
| 206 |
+
"lobster": 71,
|
| 207 |
+
"microphone": 72,
|
| 208 |
+
"monkey": 73,
|
| 209 |
+
"mosquito": 74,
|
| 210 |
+
"mouse": 75,
|
| 211 |
+
"mug": 76,
|
| 212 |
+
"mushroom": 77,
|
| 213 |
+
"onion": 78,
|
| 214 |
+
"panda": 79,
|
| 215 |
+
"peanut": 80,
|
| 216 |
+
"pear": 81,
|
| 217 |
+
"peas": 82,
|
| 218 |
+
"pencil": 83,
|
| 219 |
+
"penguin": 84,
|
| 220 |
+
"pig": 85,
|
| 221 |
+
"pillow": 86,
|
| 222 |
+
"pineapple": 87,
|
| 223 |
+
"potato": 88,
|
| 224 |
+
"power_outlet": 89,
|
| 225 |
+
"purse": 90,
|
| 226 |
+
"rabbit": 91,
|
| 227 |
+
"raccoon": 92,
|
| 228 |
+
"rhinoceros": 93,
|
| 229 |
+
"rifle": 94,
|
| 230 |
+
"saxophone": 95,
|
| 231 |
+
"screwdriver": 96,
|
| 232 |
+
"sea_turtle": 97,
|
| 233 |
+
"see_saw": 98,
|
| 234 |
+
"sheep": 99,
|
| 235 |
+
"shoe": 100,
|
| 236 |
+
"skateboard": 101,
|
| 237 |
+
"snake": 102,
|
| 238 |
+
"speedboat": 103,
|
| 239 |
+
"spider": 104,
|
| 240 |
+
"squirrel": 105,
|
| 241 |
+
"strawberry": 106,
|
| 242 |
+
"streetlight": 107,
|
| 243 |
+
"string_bean": 108,
|
| 244 |
+
"submarine": 109,
|
| 245 |
+
"swan": 110,
|
| 246 |
+
"table": 111,
|
| 247 |
+
"teapot": 112,
|
| 248 |
+
"teddy-bear": 113,
|
| 249 |
+
"television": 114,
|
| 250 |
+
"the_Eiffel_Tower": 115,
|
| 251 |
+
"the_Great_Wall_of_China": 116,
|
| 252 |
+
"tiger": 117,
|
| 253 |
+
"toe": 118,
|
| 254 |
+
"train": 119,
|
| 255 |
+
"truck": 120,
|
| 256 |
+
"umbrella": 121,
|
| 257 |
+
"vase": 122,
|
| 258 |
+
"watermelon": 123,
|
| 259 |
+
"whale": 124,
|
| 260 |
+
"zebra": 125
|
| 261 |
+
},
|
| 262 |
+
"model_type": "siglip",
|
| 263 |
+
"problem_type": "single_label_classification",
|
| 264 |
+
"text_config": {
|
| 265 |
+
"attention_dropout": 0.0,
|
| 266 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 267 |
+
"hidden_size": 768,
|
| 268 |
+
"intermediate_size": 3072,
|
| 269 |
+
"layer_norm_eps": 1e-06,
|
| 270 |
+
"max_position_embeddings": 64,
|
| 271 |
+
"model_type": "siglip_text_model",
|
| 272 |
+
"num_attention_heads": 12,
|
| 273 |
+
"num_hidden_layers": 12,
|
| 274 |
+
"projection_size": 768,
|
| 275 |
+
"torch_dtype": "float32",
|
| 276 |
+
"vocab_size": 256000
|
| 277 |
+
},
|
| 278 |
+
"torch_dtype": "float32",
|
| 279 |
+
"transformers_version": "4.51.0.dev0",
|
| 280 |
+
"vision_config": {
|
| 281 |
+
"attention_dropout": 0.0,
|
| 282 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 283 |
+
"hidden_size": 768,
|
| 284 |
+
"image_size": 224,
|
| 285 |
+
"intermediate_size": 3072,
|
| 286 |
+
"layer_norm_eps": 1e-06,
|
| 287 |
+
"model_type": "siglip_vision_model",
|
| 288 |
+
"num_attention_heads": 12,
|
| 289 |
+
"num_channels": 3,
|
| 290 |
+
"num_hidden_layers": 12,
|
| 291 |
+
"patch_size": 16,
|
| 292 |
+
"torch_dtype": "float32"
|
| 293 |
+
}
|
| 294 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3c48077d6d68a4f4556e6e0343e3ad578bbcbdcf027767c3622cb6366f64e0c
|
| 3 |
+
size 371949432
|
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:d60dfc3a7f3feaaf5fab64c05e89c94efd07d977dc0d1cf7a28f5e86e8cbf7e5
|
| 3 |
+
size 5304
|