{
  "_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": "bus stop",
    "1": "tree branch",
    "2": "silver and black",
    "3": "french",
    "4": "pastries",
    "5": "security",
    "6": "blue and white",
    "7": "open",
    "8": "cage",
    "9": "plain",
    "10": "red and blue",
    "11": "soccer ball",
    "12": "17",
    "13": "20",
    "14": "beige",
    "15": "people",
    "16": "snowboard",
    "17": "13",
    "18": "suv",
    "19": "7:45",
    "20": "can't tell",
    "21": "gray and white",
    "22": "air",
    "23": "on grass",
    "24": "calico",
    "25": "clear",
    "26": "clock",
    "27": "on road",
    "28": "rain",
    "29": "happiness",
    "30": "15",
    "31": "german",
    "32": "graffiti",
    "33": "not possible",
    "34": "orange",
    "35": "exit",
    "36": "14",
    "37": "on bench",
    "38": "fashion",
    "39": "daisy",
    "40": "black",
    "41": "brown",
    "42": "ball",
    "43": "plate",
    "44": "2013",
    "45": "snowboarder",
    "46": "pelican",
    "47": "bird",
    "48": "parking lot",
    "49": "solid",
    "50": "train",
    "51": "lg",
    "52": "yellow and orange",
    "53": "in water",
    "54": "neither",
    "55": "unknown",
    "56": "2000",
    "57": "leather",
    "58": "bear",
    "59": "motorbike",
    "60": "i don't know",
    "61": "airplanes",
    "62": "daisies",
    "63": "marker",
    "64": "gas",
    "65": "surfer",
    "66": "white and blue",
    "67": "bikes",
    "68": "cloudy",
    "69": "waiting",
    "70": "doughnut",
    "71": "landing",
    "72": "looking out window",
    "73": "at table",
    "74": "1",
    "75": "human",
    "76": "resting",
    "77": "road",
    "78": "happy",
    "79": "table",
    "80": "cross",
    "81": "cleaning",
    "82": "7:35",
    "83": "harley",
    "84": "donut",
    "85": "lying down",
    "86": "plastic",
    "87": "donuts",
    "88": "2",
    "89": "green",
    "90": "stripes",
    "91": "ostrich",
    "92": "talking",
    "93": "me",
    "94": "marble",
    "95": "nobody",
    "96": "10",
    "97": "left",
    "98": "real",
    "99": "bush",
    "100": "wild",
    "101": "dirt",
    "102": "africa",
    "103": "18",
    "104": "shoes",
    "105": "8",
    "106": "earring",
    "107": "roses",
    "108": "snowboarding",
    "109": "screen",
    "110": "bike",
    "111": "in field",
    "112": "down",
    "113": "not here",
    "114": "rack",
    "115": "forward",
    "116": "purse",
    "117": "hardwood",
    "118": "curtains",
    "119": "shrimp",
    "120": "morning",
    "121": "lilies",
    "122": "talking on phone",
    "123": "sad",
    "124": "giraffe",
    "125": "wood",
    "126": "wedding",
    "127": "skiing",
    "128": "tired",
    "129": "corn",
    "130": "daytime",
    "131": "lifeguard",
    "132": "fence",
    "133": "plane",
    "134": "lots",
    "135": "very",
    "136": "dusk",
    "137": "good",
    "138": "savannah",
    "139": "lamp",
    "140": "laptop",
    "141": "computer",
    "142": "skateboarding",
    "143": "skyscraper",
    "144": "couch",
    "145": "inside",
    "146": "regular",
    "147": "white",
    "148": "3",
    "149": "washing",
    "150": "bowl",
    "151": "luggage",
    "152": "metal",
    "153": "protection",
    "154": "right",
    "155": "walking",
    "156": "skier",
    "157": "toilet",
    "158": "door",
    "159": "nowhere",
    "160": "natural",
    "161": "ears",
    "162": "usa",
    "163": "front",
    "164": "11",
    "165": "ear",
    "166": "picnic table",
    "167": "big ben",
    "168": "windows",
    "169": "baseball",
    "170": "fishing",
    "171": "chopsticks",
    "172": "hugging",
    "173": "net",
    "174": "dress",
    "175": "display",
    "176": "student",
    "177": "in car",
    "178": "king",
    "179": "9:35",
    "180": "ceramic",
    "181": "person",
    "182": "sun",
    "183": "boy",
    "184": "safari",
    "185": "6",
    "186": "hair",
    "187": "grazing",
    "188": "blonde",
    "189": "sidewalk",
    "190": "bathroom",
    "191": "photographer",
    "192": "monitor",
    "193": "wiimote",
    "194": "crown",
    "195": "queen",
    "196": "motorcycle",
    "197": "bag",
    "198": "necklace",
    "199": "giraffes",
    "200": "lady",
    "201": "button up",
    "202": "no",
    "203": "tent",
    "204": "not there",
    "205": "cat",
    "206": "not sure",
    "207": "tv",
    "208": "conductor",
    "209": "remote",
    "210": "afternoon",
    "211": "street",
    "212": "train tracks",
    "213": "carnations",
    "214": "man",
    "215": "cement",
    "216": "bicycles",
    "217": "smiling",
    "218": "beagle",
    "219": "not very",
    "220": "male",
    "221": "purple",
    "222": "serious",
    "223": "out",
    "224": "shelter",
    "225": "little girl",
    "226": "small",
    "227": "noon",
    "228": "email",
    "229": "owner",
    "230": "can't see",
    "231": "16",
    "232": "branch",
    "233": "cannot tell",
    "234": "concrete",
    "235": "don't know",
    "236": "many",
    "237": "bicycle",
    "238": "electric",
    "239": "american",
    "240": "tour",
    "241": "engine",
    "242": "dog",
    "243": "rv",
    "244": "helmet",
    "245": "turkey",
    "246": "large",
    "247": "paper",
    "248": "4",
    "249": "tower",
    "250": "wine tasting",
    "251": "color",
    "252": "smile",
    "253": "umbrellas",
    "254": "trees",
    "255": "model",
    "256": "pink",
    "257": "passengers",
    "258": "white and black",
    "259": "red and yellow",
    "260": "curtain",
    "261": "women",
    "262": "stand",
    "263": "on street",
    "264": "orange and yellow",
    "265": "tulips",
    "266": "gray and black",
    "267": "sleeping",
    "268": "straight",
    "269": "television",
    "270": "sky",
    "271": "hat",
    "272": "car",
    "273": "white and green",
    "274": "bricks",
    "275": "rose",
    "276": "they aren't",
    "277": "asphalt",
    "278": "backpack",
    "279": "fire hydrant",
    "280": "bus",
    "281": "standing",
    "282": "red",
    "283": "in air",
    "284": "closed",
    "285": "canopy",
    "286": "shadow",
    "287": "crossing",
    "288": "white and brown",
    "289": "brick",
    "290": "girl",
    "291": "wine",
    "292": "style",
    "293": "female",
    "294": "8:35",
    "295": "birthday",
    "296": "bored",
    "297": "skateboard",
    "298": "forest",
    "299": "double",
    "300": "crane",
    "301": "above",
    "302": "tracks",
    "303": "motor",
    "304": "husky",
    "305": "snow",
    "306": "humans",
    "307": "suitcase",
    "308": "platform",
    "309": "blue",
    "310": "airplane",
    "311": "chicago",
    "312": "center",
    "313": "avocado",
    "314": "under",
    "315": "woods",
    "316": "soccer",
    "317": "field",
    "318": "bench",
    "319": "cubs",
    "320": "zoo",
    "321": "shadows",
    "322": "5",
    "323": "full",
    "324": "sunny",
    "325": "porcelain",
    "326": "tan",
    "327": "taking off",
    "328": "grass",
    "329": "wall",
    "330": "toilet paper",
    "331": "laying down",
    "332": "light",
    "333": "woman",
    "334": "camera",
    "335": "desert",
    "336": "bike rack",
    "337": "window",
    "338": "outside",
    "339": "name tag",
    "340": "cranes",
    "341": "2010",
    "342": "cap",
    "343": "silver",
    "344": "cream",
    "345": "ground",
    "346": "glass",
    "347": "yellow",
    "348": "tabby",
    "349": "roof",
    "350": "sofa",
    "351": "arrow",
    "352": "ice cream",
    "353": "chair",
    "354": "pelicans",
    "355": "hydrant",
    "356": "neon",
    "357": "track",
    "358": "young",
    "359": "station",
    "360": "twin",
    "361": "unsure",
    "362": "7",
    "363": "jeep",
    "364": "nothing",
    "365": "park",
    "366": "angry",
    "367": "watching",
    "368": "bedroom",
    "369": "hawaii",
    "370": "parked",
    "371": "over",
    "372": "hotel room",
    "373": "ducati",
    "374": "square",
    "375": "tree",
    "376": "yes",
    "377": "shade",
    "378": "church",
    "379": "cup",
    "380": "0",
    "381": "single",
    "382": "lanyard",
    "383": "clock tower",
    "384": "wii",
    "385": "gray",
    "386": "shower",
    "387": "boredom",
    "388": "black and white",
    "389": "low",
    "390": "raining",
    "391": "19"
  },
  "image_size": 384,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "0": 380,
    "1": 74,
    "10": 96,
    "11": 164,
    "13": 17,
    "14": 36,
    "15": 30,
    "16": 231,
    "17": 12,
    "18": 103,
    "19": 391,
    "2": 88,
    "20": 13,
    "2000": 56,
    "2010": 341,
    "2013": 44,
    "3": 148,
    "4": 248,
    "5": 322,
    "6": 185,
    "7": 362,
    "7:35": 82,
    "7:45": 19,
    "8": 105,
    "8:35": 294,
    "9:35": 179,
    "above": 301,
    "africa": 102,
    "afternoon": 210,
    "air": 22,
    "airplane": 310,
    "airplanes": 61,
    "american": 239,
    "angry": 366,
    "arrow": 351,
    "asphalt": 277,
    "at table": 73,
    "avocado": 313,
    "backpack": 278,
    "bag": 197,
    "ball": 42,
    "baseball": 169,
    "bathroom": 190,
    "beagle": 218,
    "bear": 58,
    "bedroom": 368,
    "beige": 14,
    "bench": 318,
    "bicycle": 237,
    "bicycles": 216,
    "big ben": 167,
    "bike": 110,
    "bike rack": 336,
    "bikes": 67,
    "bird": 47,
    "birthday": 295,
    "black": 40,
    "black and white": 388,
    "blonde": 188,
    "blue": 309,
    "blue and white": 6,
    "bored": 296,
    "boredom": 387,
    "bowl": 150,
    "boy": 183,
    "branch": 232,
    "brick": 289,
    "bricks": 274,
    "brown": 41,
    "bus": 280,
    "bus stop": 0,
    "bush": 99,
    "button up": 201,
    "cage": 8,
    "calico": 24,
    "camera": 334,
    "can't see": 230,
    "can't tell": 20,
    "cannot tell": 233,
    "canopy": 285,
    "cap": 342,
    "car": 272,
    "carnations": 213,
    "cat": 205,
    "cement": 215,
    "center": 312,
    "ceramic": 180,
    "chair": 353,
    "chicago": 311,
    "chopsticks": 171,
    "church": 378,
    "cleaning": 81,
    "clear": 25,
    "clock": 26,
    "clock tower": 383,
    "closed": 284,
    "cloudy": 68,
    "color": 251,
    "computer": 141,
    "concrete": 234,
    "conductor": 208,
    "corn": 129,
    "couch": 144,
    "crane": 300,
    "cranes": 340,
    "cream": 344,
    "cross": 80,
    "crossing": 287,
    "crown": 194,
    "cubs": 319,
    "cup": 379,
    "curtain": 260,
    "curtains": 118,
    "daisies": 62,
    "daisy": 39,
    "daytime": 130,
    "desert": 335,
    "dirt": 101,
    "display": 175,
    "dog": 242,
    "don't know": 235,
    "donut": 84,
    "donuts": 87,
    "door": 158,
    "double": 299,
    "doughnut": 70,
    "down": 112,
    "dress": 174,
    "ducati": 373,
    "dusk": 136,
    "ear": 165,
    "earring": 106,
    "ears": 161,
    "electric": 238,
    "email": 228,
    "engine": 241,
    "exit": 35,
    "fashion": 38,
    "female": 293,
    "fence": 132,
    "field": 317,
    "fire hydrant": 279,
    "fishing": 170,
    "forest": 298,
    "forward": 115,
    "french": 3,
    "front": 163,
    "full": 323,
    "gas": 64,
    "german": 31,
    "giraffe": 124,
    "giraffes": 199,
    "girl": 290,
    "glass": 346,
    "good": 137,
    "graffiti": 32,
    "grass": 328,
    "gray": 385,
    "gray and black": 266,
    "gray and white": 21,
    "grazing": 187,
    "green": 89,
    "ground": 345,
    "hair": 186,
    "happiness": 29,
    "happy": 78,
    "hardwood": 117,
    "harley": 83,
    "hat": 271,
    "hawaii": 369,
    "helmet": 244,
    "hotel room": 372,
    "hugging": 172,
    "human": 75,
    "humans": 306,
    "husky": 304,
    "hydrant": 355,
    "i don't know": 60,
    "ice cream": 352,
    "in air": 283,
    "in car": 177,
    "in field": 111,
    "in water": 53,
    "inside": 145,
    "jeep": 363,
    "king": 178,
    "lady": 200,
    "lamp": 139,
    "landing": 71,
    "lanyard": 382,
    "laptop": 140,
    "large": 246,
    "laying down": 331,
    "leather": 57,
    "left": 97,
    "lg": 51,
    "lifeguard": 131,
    "light": 332,
    "lilies": 121,
    "little girl": 225,
    "looking out window": 72,
    "lots": 134,
    "low": 389,
    "luggage": 151,
    "lying down": 85,
    "male": 220,
    "man": 214,
    "many": 236,
    "marble": 94,
    "marker": 63,
    "me": 93,
    "metal": 152,
    "model": 255,
    "monitor": 192,
    "morning": 120,
    "motor": 303,
    "motorbike": 59,
    "motorcycle": 196,
    "name tag": 339,
    "natural": 160,
    "necklace": 198,
    "neither": 54,
    "neon": 356,
    "net": 173,
    "no": 202,
    "nobody": 95,
    "noon": 227,
    "not here": 113,
    "not possible": 33,
    "not sure": 206,
    "not there": 204,
    "not very": 219,
    "nothing": 364,
    "nowhere": 159,
    "on bench": 37,
    "on grass": 23,
    "on road": 27,
    "on street": 263,
    "open": 7,
    "orange": 34,
    "orange and yellow": 264,
    "ostrich": 91,
    "out": 223,
    "outside": 338,
    "over": 371,
    "owner": 229,
    "paper": 247,
    "park": 365,
    "parked": 370,
    "parking lot": 48,
    "passengers": 257,
    "pastries": 4,
    "pelican": 46,
    "pelicans": 354,
    "people": 15,
    "person": 181,
    "photographer": 191,
    "picnic table": 166,
    "pink": 256,
    "plain": 9,
    "plane": 133,
    "plastic": 86,
    "plate": 43,
    "platform": 308,
    "porcelain": 325,
    "protection": 153,
    "purple": 221,
    "purse": 116,
    "queen": 195,
    "rack": 114,
    "rain": 28,
    "raining": 390,
    "real": 98,
    "red": 282,
    "red and blue": 10,
    "red and yellow": 259,
    "regular": 146,
    "remote": 209,
    "resting": 76,
    "right": 154,
    "road": 77,
    "roof": 349,
    "rose": 275,
    "roses": 107,
    "rv": 243,
    "sad": 123,
    "safari": 184,
    "savannah": 138,
    "screen": 109,
    "security": 5,
    "serious": 222,
    "shade": 377,
    "shadow": 286,
    "shadows": 321,
    "shelter": 224,
    "shoes": 104,
    "shower": 386,
    "shrimp": 119,
    "sidewalk": 189,
    "silver": 343,
    "silver and black": 2,
    "single": 381,
    "skateboard": 297,
    "skateboarding": 142,
    "skier": 156,
    "skiing": 127,
    "sky": 270,
    "skyscraper": 143,
    "sleeping": 267,
    "small": 226,
    "smile": 252,
    "smiling": 217,
    "snow": 305,
    "snowboard": 16,
    "snowboarder": 45,
    "snowboarding": 108,
    "soccer": 316,
    "soccer ball": 11,
    "sofa": 350,
    "solid": 49,
    "square": 374,
    "stand": 262,
    "standing": 281,
    "station": 359,
    "straight": 268,
    "street": 211,
    "stripes": 90,
    "student": 176,
    "style": 292,
    "suitcase": 307,
    "sun": 182,
    "sunny": 324,
    "surfer": 65,
    "suv": 18,
    "tabby": 348,
    "table": 79,
    "taking off": 327,
    "talking": 92,
    "talking on phone": 122,
    "tan": 326,
    "television": 269,
    "tent": 203,
    "they aren't": 276,
    "tired": 128,
    "toilet": 157,
    "toilet paper": 330,
    "tour": 240,
    "tower": 249,
    "track": 357,
    "tracks": 302,
    "train": 50,
    "train tracks": 212,
    "tree": 375,
    "tree branch": 1,
    "trees": 254,
    "tulips": 265,
    "turkey": 245,
    "tv": 207,
    "twin": 360,
    "umbrellas": 253,
    "under": 314,
    "unknown": 55,
    "unsure": 361,
    "usa": 162,
    "very": 135,
    "waiting": 69,
    "walking": 155,
    "wall": 329,
    "washing": 149,
    "watching": 367,
    "wedding": 126,
    "white": 147,
    "white and black": 258,
    "white and blue": 66,
    "white and brown": 288,
    "white and green": 273,
    "wii": 384,
    "wiimote": 193,
    "wild": 100,
    "window": 337,
    "windows": 168,
    "wine": 291,
    "wine tasting": 250,
    "woman": 333,
    "women": 261,
    "wood": 125,
    "woods": 315,
    "yellow": 347,
    "yellow and orange": 52,
    "yes": 376,
    "young": 358,
    "zoo": 320
  },
  "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": "4.29.2",
  "type_vocab_size": 2,
  "vocab_size": 30522
}