vilt_finetuned_200 / config.json
hilariooliveira's picture
End of training
b510437 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": "french",
"1": "plain",
"2": "arrow",
"3": "green",
"4": "skateboard",
"5": "down",
"6": "protection",
"7": "lady",
"8": "snowboard",
"9": "black",
"10": "lanyard",
"11": "cross",
"12": "style",
"13": "man",
"14": "don't know",
"15": "africa",
"16": "4",
"17": "hawaii",
"18": "boy",
"19": "sidewalk",
"20": "skier",
"21": "walking",
"22": "8:35",
"23": "king",
"24": "door",
"25": "bricks",
"26": "security",
"27": "church",
"28": "window",
"29": "gray",
"30": "skiing",
"31": "white",
"32": "red and yellow",
"33": "tv",
"34": "canopy",
"35": "cup",
"36": "giraffes",
"37": "red and blue",
"38": "camera",
"39": "shrimp",
"40": "blue",
"41": "suv",
"42": "crossing",
"43": "in car",
"44": "roof",
"45": "talking",
"46": "queen",
"47": "on road",
"48": "shelter",
"49": "women",
"50": "tan",
"51": "snow",
"52": "smiling",
"53": "bike rack",
"54": "black and white",
"55": "fence",
"56": "smile",
"57": "5",
"58": "sleeping",
"59": "table",
"60": "exit",
"61": "picnic table",
"62": "dog",
"63": "windows",
"64": "right",
"65": "purple",
"66": "many",
"67": "blonde",
"68": "woman",
"69": "calico",
"70": "7",
"71": "monitor",
"72": "0",
"73": "low",
"74": "zoo",
"75": "tired",
"76": "street",
"77": "person",
"78": "yes",
"79": "pink",
"80": "double",
"81": "clear",
"82": "gray and black",
"83": "natural",
"84": "ground",
"85": "lg",
"86": "can't tell",
"87": "clock tower",
"88": "shadow",
"89": "red",
"90": "blue and white",
"91": "soccer",
"92": "girl",
"93": "bikes",
"94": "air",
"95": "beagle",
"96": "soccer ball",
"97": "nothing",
"98": "screen",
"99": "stand",
"100": "not sure",
"101": "orange",
"102": "snowboarding",
"103": "2013",
"104": "shadows",
"105": "laying down",
"106": "wine tasting",
"107": "photographer",
"108": "happy",
"109": "yellow",
"110": "bedroom",
"111": "solid",
"112": "wall",
"113": "station",
"114": "ball",
"115": "hair",
"116": "sky",
"117": "crown",
"118": "2010",
"119": "not there",
"120": "lying down",
"121": "plastic",
"122": "large",
"123": "white and black",
"124": "cloudy",
"125": "6",
"126": "platform",
"127": "they aren't",
"128": "curtains",
"129": "birthday",
"130": "hat",
"131": "neon",
"132": "snowboarder",
"133": "tower",
"134": "watching",
"135": "full",
"136": "backpack",
"137": "doughnut",
"138": "cage",
"139": "sun",
"140": "1",
"141": "resting",
"142": "trees",
"143": "donut",
"144": "train",
"145": "talking on phone",
"146": "bicycles",
"147": "beige",
"148": "9:35",
"149": "jeep",
"150": "clock",
"151": "name tag",
"152": "ice cream",
"153": "white and blue",
"154": "bus",
"155": "3",
"156": "brown",
"157": "outside",
"158": "giraffe",
"159": "out",
"160": "chair",
"161": "big ben",
"162": "small",
"163": "bicycle",
"164": "2000",
"165": "woods",
"166": "unknown",
"167": "shade",
"168": "net",
"169": "desert",
"170": "little girl",
"171": "fashion",
"172": "cat",
"173": "2",
"174": "tabby",
"175": "leather",
"176": "7:45",
"177": "chopsticks",
"178": "on street",
"179": "no",
"180": "rack",
"181": "park",
"182": "stripes",
"183": "skateboarding",
"184": "dirt",
"185": "car",
"186": "wedding",
"187": "necklace",
"188": "curtain",
"189": "10",
"190": "at table",
"191": "forest",
"192": "plate",
"193": "human",
"194": "brick",
"195": "tent",
"196": "8",
"197": "7:35",
"198": "wine"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 72,
"1": 140,
"10": 189,
"2": 173,
"2000": 164,
"2010": 118,
"2013": 103,
"3": 155,
"4": 16,
"5": 57,
"6": 125,
"7": 70,
"7:35": 197,
"7:45": 176,
"8": 196,
"8:35": 22,
"9:35": 148,
"africa": 15,
"air": 94,
"arrow": 2,
"at table": 190,
"backpack": 136,
"ball": 114,
"beagle": 95,
"bedroom": 110,
"beige": 147,
"bicycle": 163,
"bicycles": 146,
"big ben": 161,
"bike rack": 53,
"bikes": 93,
"birthday": 129,
"black": 9,
"black and white": 54,
"blonde": 67,
"blue": 40,
"blue and white": 90,
"boy": 18,
"brick": 194,
"bricks": 25,
"brown": 156,
"bus": 154,
"cage": 138,
"calico": 69,
"camera": 38,
"can't tell": 86,
"canopy": 34,
"car": 185,
"cat": 172,
"chair": 160,
"chopsticks": 177,
"church": 27,
"clear": 81,
"clock": 150,
"clock tower": 87,
"cloudy": 124,
"cross": 11,
"crossing": 42,
"crown": 117,
"cup": 35,
"curtain": 188,
"curtains": 128,
"desert": 169,
"dirt": 184,
"dog": 62,
"don't know": 14,
"donut": 143,
"door": 24,
"double": 80,
"doughnut": 137,
"down": 5,
"exit": 60,
"fashion": 171,
"fence": 55,
"forest": 191,
"french": 0,
"full": 135,
"giraffe": 158,
"giraffes": 36,
"girl": 92,
"gray": 29,
"gray and black": 82,
"green": 3,
"ground": 84,
"hair": 115,
"happy": 108,
"hat": 130,
"hawaii": 17,
"human": 193,
"ice cream": 152,
"in car": 43,
"jeep": 149,
"king": 23,
"lady": 7,
"lanyard": 10,
"large": 122,
"laying down": 105,
"leather": 175,
"lg": 85,
"little girl": 170,
"low": 73,
"lying down": 120,
"man": 13,
"many": 66,
"monitor": 71,
"name tag": 151,
"natural": 83,
"necklace": 187,
"neon": 131,
"net": 168,
"no": 179,
"not sure": 100,
"not there": 119,
"nothing": 97,
"on road": 47,
"on street": 178,
"orange": 101,
"out": 159,
"outside": 157,
"park": 181,
"person": 77,
"photographer": 107,
"picnic table": 61,
"pink": 79,
"plain": 1,
"plastic": 121,
"plate": 192,
"platform": 126,
"protection": 6,
"purple": 65,
"queen": 46,
"rack": 180,
"red": 89,
"red and blue": 37,
"red and yellow": 32,
"resting": 141,
"right": 64,
"roof": 44,
"screen": 98,
"security": 26,
"shade": 167,
"shadow": 88,
"shadows": 104,
"shelter": 48,
"shrimp": 39,
"sidewalk": 19,
"skateboard": 4,
"skateboarding": 183,
"skier": 20,
"skiing": 30,
"sky": 116,
"sleeping": 58,
"small": 162,
"smile": 56,
"smiling": 52,
"snow": 51,
"snowboard": 8,
"snowboarder": 132,
"snowboarding": 102,
"soccer": 91,
"soccer ball": 96,
"solid": 111,
"stand": 99,
"station": 113,
"street": 76,
"stripes": 182,
"style": 12,
"sun": 139,
"suv": 41,
"tabby": 174,
"table": 59,
"talking": 45,
"talking on phone": 145,
"tan": 50,
"tent": 195,
"they aren't": 127,
"tired": 75,
"tower": 133,
"train": 144,
"trees": 142,
"tv": 33,
"unknown": 166,
"walking": 21,
"wall": 112,
"watching": 134,
"wedding": 186,
"white": 31,
"white and black": 123,
"white and blue": 153,
"window": 28,
"windows": 63,
"wine": 198,
"wine tasting": 106,
"woman": 68,
"women": 49,
"woods": 165,
"yellow": 109,
"yes": 78,
"zoo": 74
},
"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": "4.40.2",
"type_vocab_size": 2,
"vocab_size": 30522
}