vilt_finetuned_200 / config.json
marcelo-dalmeida's picture
Training in progress, step 200
6a7e56e
{
"_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": "tabby",
"1": "shade",
"2": "tv",
"3": "fence",
"4": "shelter",
"5": "out",
"6": "unknown",
"7": "smile",
"8": "zoo",
"9": "beige",
"10": "7:45",
"11": "they aren't",
"12": "security",
"13": "not sure",
"14": "green",
"15": "double",
"16": "ice cream",
"17": "sky",
"18": "trees",
"19": "car",
"20": "3",
"21": "bicycle",
"22": "white and blue",
"23": "chair",
"24": "walking",
"25": "queen",
"26": "low",
"27": "in car",
"28": "1",
"29": "fashion",
"30": "smiling",
"31": "gray",
"32": "2000",
"33": "bus",
"34": "pink",
"35": "outside",
"36": "lying down",
"37": "human",
"38": "yellow",
"39": "on road",
"40": "roof",
"41": "name tag",
"42": "5",
"43": "ball",
"44": "skier",
"45": "station",
"46": "cup",
"47": "bricks",
"48": "cat",
"49": "hair",
"50": "solid",
"51": "neon",
"52": "2013",
"53": "cloudy",
"54": "6",
"55": "clear",
"56": "leather",
"57": "on street",
"58": "lady",
"59": "tired",
"60": "lg",
"61": "sun",
"62": "big ben",
"63": "dirt",
"64": "black and white",
"65": "9:35",
"66": "church",
"67": "plain",
"68": "hawaii",
"69": "french",
"70": "soccer",
"71": "red and blue",
"72": "at table",
"73": "skateboarding",
"74": "4",
"75": "hat",
"76": "women",
"77": "orange",
"78": "don't know",
"79": "picnic table",
"80": "white",
"81": "clock tower",
"82": "lanyard",
"83": "necklace",
"84": "beagle",
"85": "wine tasting",
"86": "canopy",
"87": "laying down",
"88": "brick",
"89": "sidewalk",
"90": "cage",
"91": "right",
"92": "no",
"93": "shrimp",
"94": "backpack",
"95": "8:35",
"96": "7:35",
"97": "plate",
"98": "net",
"99": "platform",
"100": "exit",
"101": "purple",
"102": "many",
"103": "bikes",
"104": "train",
"105": "park",
"106": "2010",
"107": "style",
"108": "snowboard",
"109": "man",
"110": "giraffe",
"111": "woman",
"112": "down",
"113": "skateboard",
"114": "brown",
"115": "person",
"116": "shadow",
"117": "bicycles",
"118": "air",
"119": "gray and black",
"120": "skiing",
"121": "dog",
"122": "jeep",
"123": "tan",
"124": "blue",
"125": "blonde",
"126": "large",
"127": "plastic",
"128": "boy",
"129": "chopsticks",
"130": "shadows",
"131": "wedding",
"132": "donut",
"133": "rack",
"134": "wall",
"135": "screen",
"136": "protection",
"137": "natural",
"138": "bike rack",
"139": "suv",
"140": "arrow",
"141": "street",
"142": "calico",
"143": "resting",
"144": "tent",
"145": "curtain",
"146": "cross",
"147": "table",
"148": "little girl",
"149": "10",
"150": "red",
"151": "snowboarding",
"152": "birthday",
"153": "black",
"154": "doughnut",
"155": "soccer ball",
"156": "stripes",
"157": "snow",
"158": "photographer",
"159": "windows",
"160": "talking",
"161": "0",
"162": "wine",
"163": "ground",
"164": "blue and white",
"165": "monitor",
"166": "stand",
"167": "africa",
"168": "white and black",
"169": "tower",
"170": "crown",
"171": "not there",
"172": "happy",
"173": "can't tell",
"174": "camera",
"175": "red and yellow",
"176": "desert",
"177": "full",
"178": "girl",
"179": "woods",
"180": "watching",
"181": "7",
"182": "nothing",
"183": "king",
"184": "forest",
"185": "8",
"186": "door",
"187": "talking on phone",
"188": "window",
"189": "crossing",
"190": "yes",
"191": "giraffes",
"192": "clock",
"193": "2",
"194": "small",
"195": "sleeping",
"196": "curtains",
"197": "snowboarder",
"198": "bedroom"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 161,
"1": 28,
"10": 149,
"2": 193,
"2000": 32,
"2010": 106,
"2013": 52,
"3": 20,
"4": 74,
"5": 42,
"6": 54,
"7": 181,
"7:35": 96,
"7:45": 10,
"8": 185,
"8:35": 95,
"9:35": 65,
"africa": 167,
"air": 118,
"arrow": 140,
"at table": 72,
"backpack": 94,
"ball": 43,
"beagle": 84,
"bedroom": 198,
"beige": 9,
"bicycle": 21,
"bicycles": 117,
"big ben": 62,
"bike rack": 138,
"bikes": 103,
"birthday": 152,
"black": 153,
"black and white": 64,
"blonde": 125,
"blue": 124,
"blue and white": 164,
"boy": 128,
"brick": 88,
"bricks": 47,
"brown": 114,
"bus": 33,
"cage": 90,
"calico": 142,
"camera": 174,
"can't tell": 173,
"canopy": 86,
"car": 19,
"cat": 48,
"chair": 23,
"chopsticks": 129,
"church": 66,
"clear": 55,
"clock": 192,
"clock tower": 81,
"cloudy": 53,
"cross": 146,
"crossing": 189,
"crown": 170,
"cup": 46,
"curtain": 145,
"curtains": 196,
"desert": 176,
"dirt": 63,
"dog": 121,
"don't know": 78,
"donut": 132,
"door": 186,
"double": 15,
"doughnut": 154,
"down": 112,
"exit": 100,
"fashion": 29,
"fence": 3,
"forest": 184,
"french": 69,
"full": 177,
"giraffe": 110,
"giraffes": 191,
"girl": 178,
"gray": 31,
"gray and black": 119,
"green": 14,
"ground": 163,
"hair": 49,
"happy": 172,
"hat": 75,
"hawaii": 68,
"human": 37,
"ice cream": 16,
"in car": 27,
"jeep": 122,
"king": 183,
"lady": 58,
"lanyard": 82,
"large": 126,
"laying down": 87,
"leather": 56,
"lg": 60,
"little girl": 148,
"low": 26,
"lying down": 36,
"man": 109,
"many": 102,
"monitor": 165,
"name tag": 41,
"natural": 137,
"necklace": 83,
"neon": 51,
"net": 98,
"no": 92,
"not sure": 13,
"not there": 171,
"nothing": 182,
"on road": 39,
"on street": 57,
"orange": 77,
"out": 5,
"outside": 35,
"park": 105,
"person": 115,
"photographer": 158,
"picnic table": 79,
"pink": 34,
"plain": 67,
"plastic": 127,
"plate": 97,
"platform": 99,
"protection": 136,
"purple": 101,
"queen": 25,
"rack": 133,
"red": 150,
"red and blue": 71,
"red and yellow": 175,
"resting": 143,
"right": 91,
"roof": 40,
"screen": 135,
"security": 12,
"shade": 1,
"shadow": 116,
"shadows": 130,
"shelter": 4,
"shrimp": 93,
"sidewalk": 89,
"skateboard": 113,
"skateboarding": 73,
"skier": 44,
"skiing": 120,
"sky": 17,
"sleeping": 195,
"small": 194,
"smile": 7,
"smiling": 30,
"snow": 157,
"snowboard": 108,
"snowboarder": 197,
"snowboarding": 151,
"soccer": 70,
"soccer ball": 155,
"solid": 50,
"stand": 166,
"station": 45,
"street": 141,
"stripes": 156,
"style": 107,
"sun": 61,
"suv": 139,
"tabby": 0,
"table": 147,
"talking": 160,
"talking on phone": 187,
"tan": 123,
"tent": 144,
"they aren't": 11,
"tired": 59,
"tower": 169,
"train": 104,
"trees": 18,
"tv": 2,
"unknown": 6,
"walking": 24,
"wall": 134,
"watching": 180,
"wedding": 131,
"white": 80,
"white and black": 168,
"white and blue": 22,
"window": 188,
"windows": 159,
"wine": 162,
"wine tasting": 85,
"woman": 111,
"women": 76,
"woods": 179,
"yellow": 38,
"yes": 190,
"zoo": 8
},
"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": "4.34.0",
"type_vocab_size": 2,
"vocab_size": 30522
}