vilt_finetuned_200 / config.json
jiy03150's picture
Training in progress, step 200
4da2d83
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": "not there",
"1": "not sure",
"2": "snowboarding",
"3": "photographer",
"4": "africa",
"5": "arrow",
"6": "6",
"7": "skateboarding",
"8": "gray and black",
"9": "skiing",
"10": "doughnut",
"11": "crown",
"12": "park",
"13": "bus",
"14": "clock tower",
"15": "natural",
"16": "ice cream",
"17": "net",
"18": "nothing",
"19": "donut",
"20": "birthday",
"21": "brick",
"22": "security",
"23": "smiling",
"24": "1",
"25": "hawaii",
"26": "curtains",
"27": "man",
"28": "trees",
"29": "leather",
"30": "street",
"31": "smile",
"32": "talking",
"33": "protection",
"34": "0",
"35": "woods",
"36": "black",
"37": "white",
"38": "jeep",
"39": "on road",
"40": "clock",
"41": "canopy",
"42": "bricks",
"43": "table",
"44": "crossing",
"45": "cat",
"46": "car",
"47": "chair",
"48": "lanyard",
"49": "person",
"50": "no",
"51": "low",
"52": "camera",
"53": "sky",
"54": "small",
"55": "pink",
"56": "exit",
"57": "blue",
"58": "plastic",
"59": "stand",
"60": "bicycles",
"61": "dirt",
"62": "snowboard",
"63": "air",
"64": "8",
"65": "zoo",
"66": "bedroom",
"67": "large",
"68": "sleeping",
"69": "tent",
"70": "laying down",
"71": "at table",
"72": "boy",
"73": "blonde",
"74": "skateboard",
"75": "cup",
"76": "yes",
"77": "door",
"78": "out",
"79": "7:35",
"80": "ball",
"81": "in car",
"82": "tired",
"83": "happy",
"84": "lady",
"85": "monitor",
"86": "unknown",
"87": "skier",
"88": "down",
"89": "windows",
"90": "gray",
"91": "ground",
"92": "backpack",
"93": "outside",
"94": "green",
"95": "wedding",
"96": "wall",
"97": "shade",
"98": "9:35",
"99": "yellow",
"100": "2000",
"101": "forest",
"102": "they aren't",
"103": "queen",
"104": "tv",
"105": "red and yellow",
"106": "brown",
"107": "desert",
"108": "tan",
"109": "sidewalk",
"110": "chopsticks",
"111": "8:35",
"112": "dog",
"113": "black and white",
"114": "many",
"115": "shadows",
"116": "french",
"117": "2013",
"118": "on street",
"119": "hair",
"120": "red",
"121": "name tag",
"122": "necklace",
"123": "little girl",
"124": "calico",
"125": "snow",
"126": "platform",
"127": "watching",
"128": "can't tell",
"129": "curtain",
"130": "2",
"131": "right",
"132": "beagle",
"133": "resting",
"134": "soccer",
"135": "roof",
"136": "style",
"137": "lying down",
"138": "full",
"139": "bikes",
"140": "7:45",
"141": "snowboarder",
"142": "big ben",
"143": "5",
"144": "train",
"145": "orange",
"146": "4",
"147": "clear",
"148": "fence",
"149": "double",
"150": "stripes",
"151": "king",
"152": "suv",
"153": "plain",
"154": "station",
"155": "wine tasting",
"156": "fashion",
"157": "human",
"158": "shadow",
"159": "giraffe",
"160": "sun",
"161": "church",
"162": "cage",
"163": "bicycle",
"164": "cross",
"165": "shrimp",
"166": "plate",
"167": "giraffes",
"168": "shelter",
"169": "lg",
"170": "women",
"171": "bike rack",
"172": "window",
"173": "red and blue",
"174": "7",
"175": "soccer ball",
"176": "rack",
"177": "white and blue",
"178": "2010",
"179": "beige",
"180": "screen",
"181": "talking on phone",
"182": "walking",
"183": "girl",
"184": "3",
"185": "tabby",
"186": "10",
"187": "neon",
"188": "cloudy",
"189": "woman",
"190": "tower",
"191": "purple",
"192": "picnic table",
"193": "hat",
"194": "white and black",
"195": "solid",
"196": "wine",
"197": "don't know",
"198": "blue and white"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 34,
"1": 24,
"10": 186,
"2": 130,
"2000": 100,
"2010": 178,
"2013": 117,
"3": 184,
"4": 146,
"5": 143,
"6": 6,
"7": 174,
"7:35": 79,
"7:45": 140,
"8": 64,
"8:35": 111,
"9:35": 98,
"africa": 4,
"air": 63,
"arrow": 5,
"at table": 71,
"backpack": 92,
"ball": 80,
"beagle": 132,
"bedroom": 66,
"beige": 179,
"bicycle": 163,
"bicycles": 60,
"big ben": 142,
"bike rack": 171,
"bikes": 139,
"birthday": 20,
"black": 36,
"black and white": 113,
"blonde": 73,
"blue": 57,
"blue and white": 198,
"boy": 72,
"brick": 21,
"bricks": 42,
"brown": 106,
"bus": 13,
"cage": 162,
"calico": 124,
"camera": 52,
"can't tell": 128,
"canopy": 41,
"car": 46,
"cat": 45,
"chair": 47,
"chopsticks": 110,
"church": 161,
"clear": 147,
"clock": 40,
"clock tower": 14,
"cloudy": 188,
"cross": 164,
"crossing": 44,
"crown": 11,
"cup": 75,
"curtain": 129,
"curtains": 26,
"desert": 107,
"dirt": 61,
"dog": 112,
"don't know": 197,
"donut": 19,
"door": 77,
"double": 149,
"doughnut": 10,
"down": 88,
"exit": 56,
"fashion": 156,
"fence": 148,
"forest": 101,
"french": 116,
"full": 138,
"giraffe": 159,
"giraffes": 167,
"girl": 183,
"gray": 90,
"gray and black": 8,
"green": 94,
"ground": 91,
"hair": 119,
"happy": 83,
"hat": 193,
"hawaii": 25,
"human": 157,
"ice cream": 16,
"in car": 81,
"jeep": 38,
"king": 151,
"lady": 84,
"lanyard": 48,
"large": 67,
"laying down": 70,
"leather": 29,
"lg": 169,
"little girl": 123,
"low": 51,
"lying down": 137,
"man": 27,
"many": 114,
"monitor": 85,
"name tag": 121,
"natural": 15,
"necklace": 122,
"neon": 187,
"net": 17,
"no": 50,
"not sure": 1,
"not there": 0,
"nothing": 18,
"on road": 39,
"on street": 118,
"orange": 145,
"out": 78,
"outside": 93,
"park": 12,
"person": 49,
"photographer": 3,
"picnic table": 192,
"pink": 55,
"plain": 153,
"plastic": 58,
"plate": 166,
"platform": 126,
"protection": 33,
"purple": 191,
"queen": 103,
"rack": 176,
"red": 120,
"red and blue": 173,
"red and yellow": 105,
"resting": 133,
"right": 131,
"roof": 135,
"screen": 180,
"security": 22,
"shade": 97,
"shadow": 158,
"shadows": 115,
"shelter": 168,
"shrimp": 165,
"sidewalk": 109,
"skateboard": 74,
"skateboarding": 7,
"skier": 87,
"skiing": 9,
"sky": 53,
"sleeping": 68,
"small": 54,
"smile": 31,
"smiling": 23,
"snow": 125,
"snowboard": 62,
"snowboarder": 141,
"snowboarding": 2,
"soccer": 134,
"soccer ball": 175,
"solid": 195,
"stand": 59,
"station": 154,
"street": 30,
"stripes": 150,
"style": 136,
"sun": 160,
"suv": 152,
"tabby": 185,
"table": 43,
"talking": 32,
"talking on phone": 181,
"tan": 108,
"tent": 69,
"they aren't": 102,
"tired": 82,
"tower": 190,
"train": 144,
"trees": 28,
"tv": 104,
"unknown": 86,
"walking": 182,
"wall": 96,
"watching": 127,
"wedding": 95,
"white": 37,
"white and black": 194,
"white and blue": 177,
"window": 172,
"windows": 89,
"wine": 196,
"wine tasting": 155,
"woman": 189,
"women": 170,
"woods": 35,
"yellow": 99,
"yes": 76,
"zoo": 65
},
"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
}