vilt_finetuned_200 / config.json
MahimaNR's picture
Training in progress, step 200
0d44d09 verified
raw
history blame
No virus
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": "street",
"1": "suv",
"2": "purple",
"3": "snow",
"4": "large",
"5": "happy",
"6": "necklace",
"7": "red",
"8": "trees",
"9": "ground",
"10": "black",
"11": "white",
"12": "1",
"13": "style",
"14": "sun",
"15": "7:45",
"16": "bicycle",
"17": "soccer ball",
"18": "dog",
"19": "table",
"20": "boy",
"21": "sidewalk",
"22": "rack",
"23": "nothing",
"24": "window",
"25": "air",
"26": "many",
"27": "stripes",
"28": "women",
"29": "shrimp",
"30": "2000",
"31": "big ben",
"32": "snowboard",
"33": "skier",
"34": "shadow",
"35": "plastic",
"36": "curtain",
"37": "security",
"38": "lanyard",
"39": "low",
"40": "lady",
"41": "church",
"42": "tan",
"43": "name tag",
"44": "small",
"45": "10",
"46": "yes",
"47": "on street",
"48": "sky",
"49": "tower",
"50": "8:35",
"51": "roof",
"52": "bedroom",
"53": "laying down",
"54": "protection",
"55": "3",
"56": "out",
"57": "cloudy",
"58": "french",
"59": "skateboard",
"60": "9:35",
"61": "neon",
"62": "plate",
"63": "cup",
"64": "unknown",
"65": "exit",
"66": "calico",
"67": "human",
"68": "white and blue",
"69": "station",
"70": "shade",
"71": "arrow",
"72": "talking on phone",
"73": "bikes",
"74": "picnic table",
"75": "wedding",
"76": "train",
"77": "0",
"78": "girl",
"79": "backpack",
"80": "bricks",
"81": "birthday",
"82": "natural",
"83": "brick",
"84": "ball",
"85": "curtains",
"86": "park",
"87": "tabby",
"88": "hat",
"89": "right",
"90": "net",
"91": "crossing",
"92": "6",
"93": "5",
"94": "plain",
"95": "lg",
"96": "clock tower",
"97": "hawaii",
"98": "gray and black",
"99": "little girl",
"100": "blue and white",
"101": "forest",
"102": "ice cream",
"103": "yellow",
"104": "door",
"105": "soccer",
"106": "queen",
"107": "beige",
"108": "screen",
"109": "red and yellow",
"110": "watching",
"111": "africa",
"112": "no",
"113": "2013",
"114": "talking",
"115": "fence",
"116": "clear",
"117": "stand",
"118": "hair",
"119": "woods",
"120": "snowboarding",
"121": "2",
"122": "not sure",
"123": "dirt",
"124": "tent",
"125": "green",
"126": "platform",
"127": "person",
"128": "orange",
"129": "leather",
"130": "wine tasting",
"131": "windows",
"132": "beagle",
"133": "king",
"134": "can't tell",
"135": "giraffes",
"136": "crown",
"137": "8",
"138": "bicycles",
"139": "zoo",
"140": "double",
"141": "walking",
"142": "chair",
"143": "desert",
"144": "red and blue",
"145": "fashion",
"146": "blonde",
"147": "car",
"148": "shelter",
"149": "4",
"150": "jeep",
"151": "wall",
"152": "in car",
"153": "they aren't",
"154": "bus",
"155": "bike rack",
"156": "down",
"157": "don't know",
"158": "smiling",
"159": "at table",
"160": "shadows",
"161": "on road",
"162": "man",
"163": "7:35",
"164": "full",
"165": "tired",
"166": "canopy",
"167": "cat",
"168": "tv",
"169": "skiing",
"170": "photographer",
"171": "not there",
"172": "skateboarding",
"173": "doughnut",
"174": "resting",
"175": "brown",
"176": "black and white",
"177": "solid",
"178": "cross",
"179": "clock",
"180": "woman",
"181": "monitor",
"182": "2010",
"183": "giraffe",
"184": "smile",
"185": "chopsticks",
"186": "white and black",
"187": "wine",
"188": "lying down",
"189": "gray",
"190": "cage",
"191": "pink",
"192": "7",
"193": "donut",
"194": "camera",
"195": "outside",
"196": "blue",
"197": "snowboarder",
"198": "sleeping"
},
"image_size": 384,
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"0": 77,
"1": 12,
"10": 45,
"2": 121,
"2000": 30,
"2010": 182,
"2013": 113,
"3": 55,
"4": 149,
"5": 93,
"6": 92,
"7": 192,
"7:35": 163,
"7:45": 15,
"8": 137,
"8:35": 50,
"9:35": 60,
"africa": 111,
"air": 25,
"arrow": 71,
"at table": 159,
"backpack": 79,
"ball": 84,
"beagle": 132,
"bedroom": 52,
"beige": 107,
"bicycle": 16,
"bicycles": 138,
"big ben": 31,
"bike rack": 155,
"bikes": 73,
"birthday": 81,
"black": 10,
"black and white": 176,
"blonde": 146,
"blue": 196,
"blue and white": 100,
"boy": 20,
"brick": 83,
"bricks": 80,
"brown": 175,
"bus": 154,
"cage": 190,
"calico": 66,
"camera": 194,
"can't tell": 134,
"canopy": 166,
"car": 147,
"cat": 167,
"chair": 142,
"chopsticks": 185,
"church": 41,
"clear": 116,
"clock": 179,
"clock tower": 96,
"cloudy": 57,
"cross": 178,
"crossing": 91,
"crown": 136,
"cup": 63,
"curtain": 36,
"curtains": 85,
"desert": 143,
"dirt": 123,
"dog": 18,
"don't know": 157,
"donut": 193,
"door": 104,
"double": 140,
"doughnut": 173,
"down": 156,
"exit": 65,
"fashion": 145,
"fence": 115,
"forest": 101,
"french": 58,
"full": 164,
"giraffe": 183,
"giraffes": 135,
"girl": 78,
"gray": 189,
"gray and black": 98,
"green": 125,
"ground": 9,
"hair": 118,
"happy": 5,
"hat": 88,
"hawaii": 97,
"human": 67,
"ice cream": 102,
"in car": 152,
"jeep": 150,
"king": 133,
"lady": 40,
"lanyard": 38,
"large": 4,
"laying down": 53,
"leather": 129,
"lg": 95,
"little girl": 99,
"low": 39,
"lying down": 188,
"man": 162,
"many": 26,
"monitor": 181,
"name tag": 43,
"natural": 82,
"necklace": 6,
"neon": 61,
"net": 90,
"no": 112,
"not sure": 122,
"not there": 171,
"nothing": 23,
"on road": 161,
"on street": 47,
"orange": 128,
"out": 56,
"outside": 195,
"park": 86,
"person": 127,
"photographer": 170,
"picnic table": 74,
"pink": 191,
"plain": 94,
"plastic": 35,
"plate": 62,
"platform": 126,
"protection": 54,
"purple": 2,
"queen": 106,
"rack": 22,
"red": 7,
"red and blue": 144,
"red and yellow": 109,
"resting": 174,
"right": 89,
"roof": 51,
"screen": 108,
"security": 37,
"shade": 70,
"shadow": 34,
"shadows": 160,
"shelter": 148,
"shrimp": 29,
"sidewalk": 21,
"skateboard": 59,
"skateboarding": 172,
"skier": 33,
"skiing": 169,
"sky": 48,
"sleeping": 198,
"small": 44,
"smile": 184,
"smiling": 158,
"snow": 3,
"snowboard": 32,
"snowboarder": 197,
"snowboarding": 120,
"soccer": 105,
"soccer ball": 17,
"solid": 177,
"stand": 117,
"station": 69,
"street": 0,
"stripes": 27,
"style": 13,
"sun": 14,
"suv": 1,
"tabby": 87,
"table": 19,
"talking": 114,
"talking on phone": 72,
"tan": 42,
"tent": 124,
"they aren't": 153,
"tired": 165,
"tower": 49,
"train": 76,
"trees": 8,
"tv": 168,
"unknown": 64,
"walking": 141,
"wall": 151,
"watching": 110,
"wedding": 75,
"white": 11,
"white and black": 186,
"white and blue": 68,
"window": 24,
"windows": 131,
"wine": 187,
"wine tasting": 130,
"woman": 180,
"women": 28,
"woods": 119,
"yellow": 103,
"yes": 46,
"zoo": 139
},
"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": "4.42.3",
"type_vocab_size": 2,
"vocab_size": 30522
}