config.json

{
  "architectures": [
    "ViltForQuestionAnswering"
  ],
  "attention_probs_dropout_prob": 0.0,
  "dtype": "float32",
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 768,
  "id2label": {
    "0": "yellow",
    "1": "dog",
    "2": "queen",
    "3": "foodiebaker",
    "4": "foodiebakercom",
    "5": "white, brown",
    "6": "red and orange",
    "7": "plant life",
    "8": "8",
    "9": "line",
    "10": "silly hat",
    "11": "white and black",
    "12": "white with blue lettering",
    "13": "no hats",
    "14": "outside",
    "15": "taste testing",
    "16": "skateboard",
    "17": "camperdown ross st",
    "18": "black, white",
    "19": "crossing road",
    "20": "crossing it",
    "21": "blonde",
    "22": "brown",
    "23": "spectating",
    "24": "wall",
    "25": "donut",
    "26": "leaf-hat",
    "27": "out",
    "28": "ross",
    "29": "they aren't",
    "30": "panther",
    "31": "in woods",
    "32": "1",
    "33": "no, it is queen",
    "34": "shrimp",
    "35": "to keep warm",
    "36": "many",
    "37": "wine tasting",
    "38": "none",
    "39": "red",
    "40": "orange, yellow, white, black",
    "41": "jeep",
    "42": "in jungle",
    "43": "birthday",
    "44": "king",
    "45": "resting",
    "46": "sun",
    "47": "plastic",
    "48": "it's tired",
    "49": "park",
    "50": "inward",
    "51": "red/yellow",
    "52": "red orange",
    "53": "girl",
    "54": "boy",
    "55": "donut hole",
    "56": "red and blue",
    "57": "little girl",
    "58": "ross street",
    "59": "campground",
    "60": "2",
    "61": "wedding",
    "62": "sky",
    "63": "tired",
    "64": "air",
    "65": "white and blue",
    "66": "walking",
    "67": "pink",
    "68": "gray",
    "69": "down",
    "70": "hat",
    "71": "white",
    "72": "for air",
    "73": "black",
    "74": "wine sale",
    "75": "10",
    "76": "bedroom",
    "77": "doughnut",
    "78": "wine",
    "79": "boy in blue shirt",
    "80": "fashion",
    "81": "not sure",
    "82": "4",
    "83": "airflow",
    "84": "crossing",
    "85": "in wood",
    "86": "yes, just one",
    "87": "no",
    "88": "full",
    "89": "queen sized",
    "90": "right",
    "91": "6",
    "92": "at table",
    "93": "ice cream",
    "94": "to air out barn from stinky bovines",
    "95": "none are wearing hats",
    "96": "yes",
    "97": "woods",
    "98": "ross st",
    "99": "backpack",
    "100": "beige",
    "101": "0",
    "102": "crown",
    "103": "maybe",
    "104": "this is camping",
    "105": "picnic table",
    "106": "plastic, vinyl, metal",
    "107": "chair",
    "108": "5",
    "109": "train",
    "110": "chopsticks",
    "111": "ventilation",
    "112": "car",
    "113": "cat",
    "114": "7",
    "115": "off white",
    "116": "green",
    "117": "blue and white",
    "118": "suv",
    "119": "none are",
    "120": "nobody's wearing hats",
    "121": "chopsticks spoon",
    "122": "to let sunlight in",
    "123": "leather",
    "124": "3",
    "125": "outward",
    "126": "birthday boy",
    "127": "shadow",
    "128": "style",
    "129": "table",
    "130": "palm leaves",
    "131": "forest",
    "132": "provide air",
    "133": "double",
    "134": "red and yellow",
    "135": "they are not",
    "136": "alcohol show",
    "137": "bus",
    "138": "http://foodiebakercom",
    "139": "keep cow safe",
    "140": "red, orange",
    "141": "human",
    "142": "shadows",
    "143": "nobody is wearing hats",
    "144": "woman",
    "145": "hawaii",
    "146": "top picture",
    "147": "in forest",
    "148": "cup",
    "149": "watching"
  },
  "image_size": 384,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "0": 101,
    "1": 32,
    "10": 75,
    "2": 60,
    "3": 124,
    "4": 82,
    "5": 108,
    "6": 91,
    "7": 114,
    "8": 8,
    "air": 64,
    "airflow": 83,
    "alcohol show": 136,
    "at table": 92,
    "backpack": 99,
    "bedroom": 76,
    "beige": 100,
    "birthday": 43,
    "birthday boy": 126,
    "black": 73,
    "black, white": 18,
    "blonde": 21,
    "blue and white": 117,
    "boy": 54,
    "boy in blue shirt": 79,
    "brown": 22,
    "bus": 137,
    "camperdown ross st": 17,
    "campground": 59,
    "car": 112,
    "cat": 113,
    "chair": 107,
    "chopsticks": 110,
    "chopsticks spoon": 121,
    "crossing": 84,
    "crossing it": 20,
    "crossing road": 19,
    "crown": 102,
    "cup": 148,
    "dog": 1,
    "donut": 25,
    "donut hole": 55,
    "double": 133,
    "doughnut": 77,
    "down": 69,
    "fashion": 80,
    "foodiebaker": 3,
    "foodiebakercom": 4,
    "for air": 72,
    "forest": 131,
    "full": 88,
    "girl": 53,
    "gray": 68,
    "green": 116,
    "hat": 70,
    "hawaii": 145,
    "http://foodiebakercom": 138,
    "human": 141,
    "ice cream": 93,
    "in forest": 147,
    "in jungle": 42,
    "in wood": 85,
    "in woods": 31,
    "inward": 50,
    "it's tired": 48,
    "jeep": 41,
    "keep cow safe": 139,
    "king": 44,
    "leaf-hat": 26,
    "leather": 123,
    "line": 9,
    "little girl": 57,
    "many": 36,
    "maybe": 103,
    "no": 87,
    "no hats": 13,
    "no, it is queen": 33,
    "nobody is wearing hats": 143,
    "nobody's wearing hats": 120,
    "none": 38,
    "none are": 119,
    "none are wearing hats": 95,
    "not sure": 81,
    "off white": 115,
    "orange, yellow, white, black": 40,
    "out": 27,
    "outside": 14,
    "outward": 125,
    "palm leaves": 130,
    "panther": 30,
    "park": 49,
    "picnic table": 105,
    "pink": 67,
    "plant life": 7,
    "plastic": 47,
    "plastic, vinyl, metal": 106,
    "provide air": 132,
    "queen": 2,
    "queen sized": 89,
    "red": 39,
    "red and blue": 56,
    "red and orange": 6,
    "red and yellow": 134,
    "red orange": 52,
    "red, orange": 140,
    "red/yellow": 51,
    "resting": 45,
    "right": 90,
    "ross": 28,
    "ross st": 98,
    "ross street": 58,
    "shadow": 127,
    "shadows": 142,
    "shrimp": 34,
    "silly hat": 10,
    "skateboard": 16,
    "sky": 62,
    "spectating": 23,
    "style": 128,
    "sun": 46,
    "suv": 118,
    "table": 129,
    "taste testing": 15,
    "they are not": 135,
    "they aren't": 29,
    "this is camping": 104,
    "tired": 63,
    "to air out barn from stinky bovines": 94,
    "to keep warm": 35,
    "to let sunlight in": 122,
    "top picture": 146,
    "train": 109,
    "ventilation": 111,
    "walking": 66,
    "wall": 24,
    "watching": 149,
    "wedding": 61,
    "white": 71,
    "white and black": 11,
    "white and blue": 65,
    "white with blue lettering": 12,
    "white, brown": 5,
    "wine": 78,
    "wine sale": 74,
    "wine tasting": 37,
    "woman": 144,
    "woods": 97,
    "yellow": 0,
    "yes": 96,
    "yes, just one": 86
  },
  "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,
  "pad_token_id": 0,
  "patch_size": 32,
  "qkv_bias": true,
  "tie_word_embeddings": false,
  "transformers_version": "4.56.1",
  "type_vocab_size": 2,
  "vocab_size": 30522
}