vilt_finetuned_200 / config.json
Mavish's picture
Training in progress, step 200
c9b209e verified
{
"_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": "skateboard",
"1": "cloudy",
"2": "stripes",
"3": "8",
"4": "giraffes",
"5": "backpack",
"6": "king",
"7": "nothing",
"8": "pink",
"9": "arrow",
"10": "snowboard",
"11": "don't know",
"12": "7:45",
"13": "4",
"14": "talking",
"15": "train",
"16": "camera",
"17": "style",
"18": "unknown",
"19": "snowboarding",
"20": "10",
"21": "beige",
"22": "zoo",
"23": "no",
"24": "orange",
"25": "bikes",
"26": "photographer",
"27": "blue and white",
"28": "sun",
"29": "desert",
"30": "wine",
"31": "women",
"32": "suv",
"33": "white and black",
"34": "2",
"35": "picnic table",
"36": "wine tasting",
"37": "shade",
"38": "big ben",
"39": "brown",
"40": "ground",
"41": "snow",
"42": "beagle",
"43": "lying down",
"44": "birthday",
"45": "can't tell",
"46": "monitor",
"47": "little girl",
"48": "right",
"49": "hat",
"50": "2010",
"51": "girl",
"52": "wedding",
"53": "tired",
"54": "skateboarding",
"55": "cage",
"56": "platform",
"57": "they aren't",
"58": "gray and black",
"59": "in car",
"60": "red and blue",
"61": "stand",
"62": "car",
"63": "tower",
"64": "double",
"65": "black and white",
"66": "air",
"67": "5",
"68": "9:35",
"69": "not there",
"70": "bricks",
"71": "laying down",
"72": "out",
"73": "screen",
"74": "tan",
"75": "cat",
"76": "sky",
"77": "station",
"78": "soccer",
"79": "0",
"80": "yes",
"81": "exit",
"82": "cup",
"83": "talking on phone",
"84": "resting",
"85": "windows",
"86": "bike rack",
"87": "on street",
"88": "human",
"89": "dog",
"90": "2000",
"91": "person",
"92": "plate",
"93": "down",
"94": "crossing",
"95": "low",
"96": "woods",
"97": "plain",
"98": "park",
"99": "2013",
"100": "curtain",
"101": "giraffe",
"102": "snowboarder",
"103": "tent",
"104": "shelter",
"105": "woman",
"106": "neon",
"107": "blonde",
"108": "bedroom",
"109": "purple",
"110": "natural",
"111": "shadows",
"112": "door",
"113": "red and yellow",
"114": "small",
"115": "hawaii",
"116": "black",
"117": "solid",
"118": "large",
"119": "watching",
"120": "net",
"121": "leather",
"122": "donut",
"123": "cross",
"124": "soccer ball",
"125": "french",
"126": "yellow",
"127": "fence",
"128": "happy",
"129": "sidewalk",
"130": "doughnut",
"131": "1",
"132": "white and blue",
"133": "shrimp",
"134": "dirt",
"135": "lady",
"136": "queen",
"137": "not sure",
"138": "bicycle",
"139": "trees",
"140": "bus",
"141": "green",
"142": "curtains",
"143": "6",
"144": "church",
"145": "man",
"146": "calico",
"147": "street",
"148": "chopsticks",
"149": "ice cream",
"150": "3",
"151": "walking",
"152": "roof",
"153": "lanyard",
"154": "7:35",
"155": "plastic",
"156": "at table",
"157": "brick",
"158": "wall",
"159": "fashion",
"160": "canopy",
"161": "gray",
"162": "clock tower",
"163": "skier",
"164": "lg",
"165": "ball",
"166": "tv",
"167": "full",
"168": "bicycles",
"169": "sleeping",
"170": "rack",
"171": "crown",
"172": "africa",
"173": "security",
"174": "boy",
"175": "shadow",
"176": "tabby",
"177": "blue",
"178": "window",
"179": "hair",
"180": "outside",
"181": "red",
"182": "table",
"183": "clear",
"184": "8:35",
"185": "protection",
"186": "necklace",
"187": "jeep",
"188": "white",
"189": "7",
"190": "many",
"191": "skiing",
"192": "forest",
"193": "name tag",
"194": "chair",
"195": "smile",
"196": "smiling",
"197": "clock",
"198": "on road"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 79,
"1": 131,
"10": 20,
"2": 34,
"2000": 90,
"2010": 50,
"2013": 99,
"3": 150,
"4": 13,
"5": 67,
"6": 143,
"7": 189,
"7:35": 154,
"7:45": 12,
"8": 3,
"8:35": 184,
"9:35": 68,
"africa": 172,
"air": 66,
"arrow": 9,
"at table": 156,
"backpack": 5,
"ball": 165,
"beagle": 42,
"bedroom": 108,
"beige": 21,
"bicycle": 138,
"bicycles": 168,
"big ben": 38,
"bike rack": 86,
"bikes": 25,
"birthday": 44,
"black": 116,
"black and white": 65,
"blonde": 107,
"blue": 177,
"blue and white": 27,
"boy": 174,
"brick": 157,
"bricks": 70,
"brown": 39,
"bus": 140,
"cage": 55,
"calico": 146,
"camera": 16,
"can't tell": 45,
"canopy": 160,
"car": 62,
"cat": 75,
"chair": 194,
"chopsticks": 148,
"church": 144,
"clear": 183,
"clock": 197,
"clock tower": 162,
"cloudy": 1,
"cross": 123,
"crossing": 94,
"crown": 171,
"cup": 82,
"curtain": 100,
"curtains": 142,
"desert": 29,
"dirt": 134,
"dog": 89,
"don't know": 11,
"donut": 122,
"door": 112,
"double": 64,
"doughnut": 130,
"down": 93,
"exit": 81,
"fashion": 159,
"fence": 127,
"forest": 192,
"french": 125,
"full": 167,
"giraffe": 101,
"giraffes": 4,
"girl": 51,
"gray": 161,
"gray and black": 58,
"green": 141,
"ground": 40,
"hair": 179,
"happy": 128,
"hat": 49,
"hawaii": 115,
"human": 88,
"ice cream": 149,
"in car": 59,
"jeep": 187,
"king": 6,
"lady": 135,
"lanyard": 153,
"large": 118,
"laying down": 71,
"leather": 121,
"lg": 164,
"little girl": 47,
"low": 95,
"lying down": 43,
"man": 145,
"many": 190,
"monitor": 46,
"name tag": 193,
"natural": 110,
"necklace": 186,
"neon": 106,
"net": 120,
"no": 23,
"not sure": 137,
"not there": 69,
"nothing": 7,
"on road": 198,
"on street": 87,
"orange": 24,
"out": 72,
"outside": 180,
"park": 98,
"person": 91,
"photographer": 26,
"picnic table": 35,
"pink": 8,
"plain": 97,
"plastic": 155,
"plate": 92,
"platform": 56,
"protection": 185,
"purple": 109,
"queen": 136,
"rack": 170,
"red": 181,
"red and blue": 60,
"red and yellow": 113,
"resting": 84,
"right": 48,
"roof": 152,
"screen": 73,
"security": 173,
"shade": 37,
"shadow": 175,
"shadows": 111,
"shelter": 104,
"shrimp": 133,
"sidewalk": 129,
"skateboard": 0,
"skateboarding": 54,
"skier": 163,
"skiing": 191,
"sky": 76,
"sleeping": 169,
"small": 114,
"smile": 195,
"smiling": 196,
"snow": 41,
"snowboard": 10,
"snowboarder": 102,
"snowboarding": 19,
"soccer": 78,
"soccer ball": 124,
"solid": 117,
"stand": 61,
"station": 77,
"street": 147,
"stripes": 2,
"style": 17,
"sun": 28,
"suv": 32,
"tabby": 176,
"table": 182,
"talking": 14,
"talking on phone": 83,
"tan": 74,
"tent": 103,
"they aren't": 57,
"tired": 53,
"tower": 63,
"train": 15,
"trees": 139,
"tv": 166,
"unknown": 18,
"walking": 151,
"wall": 158,
"watching": 119,
"wedding": 52,
"white": 188,
"white and black": 33,
"white and blue": 132,
"window": 178,
"windows": 85,
"wine": 30,
"wine tasting": 36,
"woman": 105,
"women": 31,
"woods": 96,
"yellow": 126,
"yes": 80,
"zoo": 22
},
"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": "4.38.2",
"type_vocab_size": 2,
"vocab_size": 30522
}