vilt_finetuned_200 / config.json
bgyoo's picture
Training in progress, step 200
b9cc469
{
"_name_or_path": "dandelin/vilt-b32-mlm",
"architectures": [
"ViltForQuestionAnswering"
],
"attention_probs_dropout_prob": 0.0,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.0,
"hidden_size": 768,
"id2label": {
"0": "pink",
"1": "dog",
"2": "africa",
"3": "chair",
"4": "giraffe",
"5": "3",
"6": "double",
"7": "wine",
"8": "table",
"9": "air",
"10": "doughnut",
"11": "monitor",
"12": "beagle",
"13": "man",
"14": "resting",
"15": "clock tower",
"16": "unknown",
"17": "in car",
"18": "ice cream",
"19": "skiing",
"20": "stand",
"21": "lying down",
"22": "black and white",
"23": "photographer",
"24": "little girl",
"25": "backpack",
"26": "woods",
"27": "leather",
"28": "cat",
"29": "fashion",
"30": "0",
"31": "red and yellow",
"32": "8:35",
"33": "ground",
"34": "skateboarding",
"35": "forest",
"36": "blue and white",
"37": "bus",
"38": "shade",
"39": "yes",
"40": "skateboard",
"41": "nothing",
"42": "don't know",
"43": "snowboard",
"44": "lanyard",
"45": "bike rack",
"46": "gray and black",
"47": "woman",
"48": "snowboarding",
"49": "snowboarder",
"50": "tired",
"51": "wedding",
"52": "women",
"53": "boy",
"54": "girl",
"55": "street",
"56": "park",
"57": "roof",
"58": "person",
"59": "not sure",
"60": "wall",
"61": "station",
"62": "jeep",
"63": "human",
"64": "bikes",
"65": "7",
"66": "soccer",
"67": "picnic table",
"68": "white and blue",
"69": "sky",
"70": "happy",
"71": "yellow",
"72": "at table",
"73": "bedroom",
"74": "queen",
"75": "crossing",
"76": "7:45",
"77": "style",
"78": "smile",
"79": "blonde",
"80": "smiling",
"81": "camera",
"82": "arrow",
"83": "neon",
"84": "small",
"85": "king",
"86": "outside",
"87": "suv",
"88": "tabby",
"89": "hat",
"90": "hawaii",
"91": "hair",
"92": "tan",
"93": "2000",
"94": "birthday",
"95": "security",
"96": "2",
"97": "tower",
"98": "can't tell",
"99": "8",
"100": "net",
"101": "clear",
"102": "curtains",
"103": "white and black",
"104": "not there",
"105": "on road",
"106": "crown",
"107": "snow",
"108": "sidewalk",
"109": "clock",
"110": "talking",
"111": "talking on phone",
"112": "cage",
"113": "solid",
"114": "wine tasting",
"115": "blue",
"116": "natural",
"117": "exit",
"118": "curtain",
"119": "church",
"120": "chopsticks",
"121": "brown",
"122": "6",
"123": "4",
"124": "2013",
"125": "rack",
"126": "skier",
"127": "shadows",
"128": "large",
"129": "9:35",
"130": "plastic",
"131": "bicycles",
"132": "they aren't",
"133": "desert",
"134": "many",
"135": "windows",
"136": "calico",
"137": "black",
"138": "10",
"139": "protection",
"140": "cloudy",
"141": "beige",
"142": "on street",
"143": "purple",
"144": "laying down",
"145": "orange",
"146": "necklace",
"147": "zoo",
"148": "right",
"149": "sleeping",
"150": "red",
"151": "french",
"152": "giraffes",
"153": "canopy",
"154": "2010",
"155": "trees",
"156": "lady",
"157": "shrimp",
"158": "shelter",
"159": "walking",
"160": "tv",
"161": "green",
"162": "bricks",
"163": "cup",
"164": "donut",
"165": "plain",
"166": "door",
"167": "sun",
"168": "7:35",
"169": "window",
"170": "big ben",
"171": "dirt",
"172": "stripes",
"173": "brick",
"174": "out",
"175": "bicycle",
"176": "train",
"177": "low",
"178": "down",
"179": "car",
"180": "gray",
"181": "name tag",
"182": "plate",
"183": "full",
"184": "platform",
"185": "5",
"186": "tent",
"187": "red and blue",
"188": "no",
"189": "fence",
"190": "ball",
"191": "watching",
"192": "cross",
"193": "1",
"194": "white",
"195": "shadow",
"196": "lg",
"197": "screen",
"198": "soccer ball"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 30,
"1": 193,
"10": 138,
"2": 96,
"2000": 93,
"2010": 154,
"2013": 124,
"3": 5,
"4": 123,
"5": 185,
"6": 122,
"7": 65,
"7:35": 168,
"7:45": 76,
"8": 99,
"8:35": 32,
"9:35": 129,
"africa": 2,
"air": 9,
"arrow": 82,
"at table": 72,
"backpack": 25,
"ball": 190,
"beagle": 12,
"bedroom": 73,
"beige": 141,
"bicycle": 175,
"bicycles": 131,
"big ben": 170,
"bike rack": 45,
"bikes": 64,
"birthday": 94,
"black": 137,
"black and white": 22,
"blonde": 79,
"blue": 115,
"blue and white": 36,
"boy": 53,
"brick": 173,
"bricks": 162,
"brown": 121,
"bus": 37,
"cage": 112,
"calico": 136,
"camera": 81,
"can't tell": 98,
"canopy": 153,
"car": 179,
"cat": 28,
"chair": 3,
"chopsticks": 120,
"church": 119,
"clear": 101,
"clock": 109,
"clock tower": 15,
"cloudy": 140,
"cross": 192,
"crossing": 75,
"crown": 106,
"cup": 163,
"curtain": 118,
"curtains": 102,
"desert": 133,
"dirt": 171,
"dog": 1,
"don't know": 42,
"donut": 164,
"door": 166,
"double": 6,
"doughnut": 10,
"down": 178,
"exit": 117,
"fashion": 29,
"fence": 189,
"forest": 35,
"french": 151,
"full": 183,
"giraffe": 4,
"giraffes": 152,
"girl": 54,
"gray": 180,
"gray and black": 46,
"green": 161,
"ground": 33,
"hair": 91,
"happy": 70,
"hat": 89,
"hawaii": 90,
"human": 63,
"ice cream": 18,
"in car": 17,
"jeep": 62,
"king": 85,
"lady": 156,
"lanyard": 44,
"large": 128,
"laying down": 144,
"leather": 27,
"lg": 196,
"little girl": 24,
"low": 177,
"lying down": 21,
"man": 13,
"many": 134,
"monitor": 11,
"name tag": 181,
"natural": 116,
"necklace": 146,
"neon": 83,
"net": 100,
"no": 188,
"not sure": 59,
"not there": 104,
"nothing": 41,
"on road": 105,
"on street": 142,
"orange": 145,
"out": 174,
"outside": 86,
"park": 56,
"person": 58,
"photographer": 23,
"picnic table": 67,
"pink": 0,
"plain": 165,
"plastic": 130,
"plate": 182,
"platform": 184,
"protection": 139,
"purple": 143,
"queen": 74,
"rack": 125,
"red": 150,
"red and blue": 187,
"red and yellow": 31,
"resting": 14,
"right": 148,
"roof": 57,
"screen": 197,
"security": 95,
"shade": 38,
"shadow": 195,
"shadows": 127,
"shelter": 158,
"shrimp": 157,
"sidewalk": 108,
"skateboard": 40,
"skateboarding": 34,
"skier": 126,
"skiing": 19,
"sky": 69,
"sleeping": 149,
"small": 84,
"smile": 78,
"smiling": 80,
"snow": 107,
"snowboard": 43,
"snowboarder": 49,
"snowboarding": 48,
"soccer": 66,
"soccer ball": 198,
"solid": 113,
"stand": 20,
"station": 61,
"street": 55,
"stripes": 172,
"style": 77,
"sun": 167,
"suv": 87,
"tabby": 88,
"table": 8,
"talking": 110,
"talking on phone": 111,
"tan": 92,
"tent": 186,
"they aren't": 132,
"tired": 50,
"tower": 97,
"train": 176,
"trees": 155,
"tv": 160,
"unknown": 16,
"walking": 159,
"wall": 60,
"watching": 191,
"wedding": 51,
"white": 194,
"white and black": 103,
"white and blue": 68,
"window": 169,
"windows": 135,
"wine": 7,
"wine tasting": 114,
"woman": 47,
"women": 52,
"woods": 26,
"yellow": 71,
"yes": 39,
"zoo": 147
},
"layer_norm_eps": 1e-12,
"max_image_length": -1,
"max_position_embeddings": 40,
"modality_type_vocab_size": 2,
"model_type": "vilt",
"num_attention_heads": 12,
"num_channels": 3,
"num_hidden_layers": 12,
"num_images": -1,
"patch_size": 32,
"qkv_bias": true,
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.31.0",
"type_vocab_size": 2,
"vocab_size": 30522
}