vilt_finetuned_200 / config.json
aqachun's picture
Training in progress, step 200
586bca5
raw
history blame
8.51 kB
{
"_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": "donut",
"1": "exit",
"2": "person",
"3": "out",
"4": "bus",
"5": "don't know",
"6": "shade",
"7": "neon",
"8": "photographer",
"9": "they aren't",
"10": "2",
"11": "sun",
"12": "cage",
"13": "shelter",
"14": "blue",
"15": "net",
"16": "human",
"17": "screen",
"18": "curtain",
"19": "7:35",
"20": "chopsticks",
"21": "air",
"22": "4",
"23": "lanyard",
"24": "dog",
"25": "red",
"26": "boy",
"27": "stand",
"28": "wine",
"29": "3",
"30": "french",
"31": "2010",
"32": "yes",
"33": "not sure",
"34": "platform",
"35": "church",
"36": "shrimp",
"37": "beige",
"38": "shadows",
"39": "canopy",
"40": "dirt",
"41": "rack",
"42": "green",
"43": "tv",
"44": "hat",
"45": "at table",
"46": "giraffe",
"47": "not there",
"48": "skateboarding",
"49": "park",
"50": "resting",
"51": "pink",
"52": "clock",
"53": "can't tell",
"54": "smiling",
"55": "station",
"56": "camera",
"57": "watching",
"58": "white and blue",
"59": "little girl",
"60": "table",
"61": "sidewalk",
"62": "birthday",
"63": "blonde",
"64": "black",
"65": "tent",
"66": "door",
"67": "7:45",
"68": "crown",
"69": "happy",
"70": "windows",
"71": "hawaii",
"72": "style",
"73": "on street",
"74": "security",
"75": "bicycle",
"76": "queen",
"77": "soccer ball",
"78": "lg",
"79": "women",
"80": "black and white",
"81": "brick",
"82": "snowboarding",
"83": "natural",
"84": "picnic table",
"85": "orange",
"86": "chair",
"87": "curtains",
"88": "nothing",
"89": "in car",
"90": "snow",
"91": "suv",
"92": "crossing",
"93": "snowboarder",
"94": "bike rack",
"95": "plastic",
"96": "7",
"97": "forest",
"98": "9:35",
"99": "cat",
"100": "brown",
"101": "africa",
"102": "cross",
"103": "laying down",
"104": "woods",
"105": "woman",
"106": "protection",
"107": "backpack",
"108": "talking on phone",
"109": "clock tower",
"110": "fashion",
"111": "clear",
"112": "skateboard",
"113": "lying down",
"114": "giraffes",
"115": "big ben",
"116": "man",
"117": "8",
"118": "full",
"119": "gray",
"120": "gray and black",
"121": "low",
"122": "name tag",
"123": "down",
"124": "white and black",
"125": "outside",
"126": "tan",
"127": "talking",
"128": "double",
"129": "bicycles",
"130": "skiing",
"131": "no",
"132": "cup",
"133": "right",
"134": "walking",
"135": "trees",
"136": "2000",
"137": "10",
"138": "doughnut",
"139": "stripes",
"140": "0",
"141": "sleeping",
"142": "lady",
"143": "soccer",
"144": "solid",
"145": "bricks",
"146": "skier",
"147": "white",
"148": "king",
"149": "wedding",
"150": "street",
"151": "shadow",
"152": "arrow",
"153": "necklace",
"154": "girl",
"155": "5",
"156": "2013",
"157": "zoo",
"158": "blue and white",
"159": "yellow",
"160": "hair",
"161": "small",
"162": "red and yellow",
"163": "ice cream",
"164": "car",
"165": "red and blue",
"166": "unknown",
"167": "plain",
"168": "desert",
"169": "sky",
"170": "jeep",
"171": "wall",
"172": "roof",
"173": "bikes",
"174": "smile",
"175": "wine tasting",
"176": "cloudy",
"177": "purple",
"178": "snowboard",
"179": "plate",
"180": "beagle",
"181": "ball",
"182": "tabby",
"183": "bedroom",
"184": "1",
"185": "8:35",
"186": "train",
"187": "monitor",
"188": "ground",
"189": "large",
"190": "on road",
"191": "tower",
"192": "tired",
"193": "window",
"194": "calico",
"195": "fence",
"196": "many",
"197": "leather",
"198": "6"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 140,
"1": 184,
"10": 137,
"2": 10,
"2000": 136,
"2010": 31,
"2013": 156,
"3": 29,
"4": 22,
"5": 155,
"6": 198,
"7": 96,
"7:35": 19,
"7:45": 67,
"8": 117,
"8:35": 185,
"9:35": 98,
"africa": 101,
"air": 21,
"arrow": 152,
"at table": 45,
"backpack": 107,
"ball": 181,
"beagle": 180,
"bedroom": 183,
"beige": 37,
"bicycle": 75,
"bicycles": 129,
"big ben": 115,
"bike rack": 94,
"bikes": 173,
"birthday": 62,
"black": 64,
"black and white": 80,
"blonde": 63,
"blue": 14,
"blue and white": 158,
"boy": 26,
"brick": 81,
"bricks": 145,
"brown": 100,
"bus": 4,
"cage": 12,
"calico": 194,
"camera": 56,
"can't tell": 53,
"canopy": 39,
"car": 164,
"cat": 99,
"chair": 86,
"chopsticks": 20,
"church": 35,
"clear": 111,
"clock": 52,
"clock tower": 109,
"cloudy": 176,
"cross": 102,
"crossing": 92,
"crown": 68,
"cup": 132,
"curtain": 18,
"curtains": 87,
"desert": 168,
"dirt": 40,
"dog": 24,
"don't know": 5,
"donut": 0,
"door": 66,
"double": 128,
"doughnut": 138,
"down": 123,
"exit": 1,
"fashion": 110,
"fence": 195,
"forest": 97,
"french": 30,
"full": 118,
"giraffe": 46,
"giraffes": 114,
"girl": 154,
"gray": 119,
"gray and black": 120,
"green": 42,
"ground": 188,
"hair": 160,
"happy": 69,
"hat": 44,
"hawaii": 71,
"human": 16,
"ice cream": 163,
"in car": 89,
"jeep": 170,
"king": 148,
"lady": 142,
"lanyard": 23,
"large": 189,
"laying down": 103,
"leather": 197,
"lg": 78,
"little girl": 59,
"low": 121,
"lying down": 113,
"man": 116,
"many": 196,
"monitor": 187,
"name tag": 122,
"natural": 83,
"necklace": 153,
"neon": 7,
"net": 15,
"no": 131,
"not sure": 33,
"not there": 47,
"nothing": 88,
"on road": 190,
"on street": 73,
"orange": 85,
"out": 3,
"outside": 125,
"park": 49,
"person": 2,
"photographer": 8,
"picnic table": 84,
"pink": 51,
"plain": 167,
"plastic": 95,
"plate": 179,
"platform": 34,
"protection": 106,
"purple": 177,
"queen": 76,
"rack": 41,
"red": 25,
"red and blue": 165,
"red and yellow": 162,
"resting": 50,
"right": 133,
"roof": 172,
"screen": 17,
"security": 74,
"shade": 6,
"shadow": 151,
"shadows": 38,
"shelter": 13,
"shrimp": 36,
"sidewalk": 61,
"skateboard": 112,
"skateboarding": 48,
"skier": 146,
"skiing": 130,
"sky": 169,
"sleeping": 141,
"small": 161,
"smile": 174,
"smiling": 54,
"snow": 90,
"snowboard": 178,
"snowboarder": 93,
"snowboarding": 82,
"soccer": 143,
"soccer ball": 77,
"solid": 144,
"stand": 27,
"station": 55,
"street": 150,
"stripes": 139,
"style": 72,
"sun": 11,
"suv": 91,
"tabby": 182,
"table": 60,
"talking": 127,
"talking on phone": 108,
"tan": 126,
"tent": 65,
"they aren't": 9,
"tired": 192,
"tower": 191,
"train": 186,
"trees": 135,
"tv": 43,
"unknown": 166,
"walking": 134,
"wall": 171,
"watching": 57,
"wedding": 149,
"white": 147,
"white and black": 124,
"white and blue": 58,
"window": 193,
"windows": 70,
"wine": 28,
"wine tasting": 175,
"woman": 105,
"women": 79,
"woods": 104,
"yellow": 159,
"yes": 32,
"zoo": 157
},
"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": "4.33.1",
"type_vocab_size": 2,
"vocab_size": 30522
}