config.json

{
  "_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": "out",
    "1": "girl",
    "2": "white and black",
    "3": "shadows",
    "4": "double",
    "5": "wine tasting",
    "6": "queen",
    "7": "not sure",
    "8": "10",
    "9": "many",
    "10": "tired",
    "11": "cat",
    "12": "woman",
    "13": "park",
    "14": "suv",
    "15": "wine",
    "16": "jeep",
    "17": "hat",
    "18": "red and yellow",
    "19": "red and blue",
    "20": "1",
    "21": "at table",
    "22": "6",
    "23": "no",
    "24": "sky",
    "25": "brown",
    "26": "hawaii",
    "27": "table",
    "28": "plastic",
    "29": "cup",
    "30": "birthday",
    "31": "donut",
    "32": "crossing",
    "33": "down",
    "34": "skateboard",
    "35": "3",
    "36": "full",
    "37": "gray",
    "38": "style",
    "39": "bedroom",
    "40": "8",
    "41": "little girl",
    "42": "pink",
    "43": "2",
    "44": "sun",
    "45": "boy",
    "46": "red",
    "47": "shadow",
    "48": "wedding",
    "49": "shrimp",
    "50": "dog",
    "51": "crown",
    "52": "king",
    "53": "bus",
    "54": "0",
    "55": "blue and white",
    "56": "wall",
    "57": "4",
    "58": "5",
    "59": "7",
    "60": "black",
    "61": "chair",
    "62": "yellow",
    "63": "car",
    "64": "train",
    "65": "fashion",
    "66": "resting",
    "67": "walking",
    "68": "woods",
    "69": "they aren't",
    "70": "backpack",
    "71": "chopsticks",
    "72": "forest",
    "73": "leather",
    "74": "white and blue",
    "75": "picnic table",
    "76": "ice cream",
    "77": "blonde",
    "78": "human",
    "79": "beige",
    "80": "green",
    "81": "air",
    "82": "doughnut",
    "83": "yes",
    "84": "right",
    "85": "watching",
    "86": "outside",
    "87": "white"
  },
  "image_size": 384,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "0": 54,
    "1": 20,
    "10": 8,
    "2": 43,
    "3": 35,
    "4": 57,
    "5": 58,
    "6": 22,
    "7": 59,
    "8": 40,
    "air": 81,
    "at table": 21,
    "backpack": 70,
    "bedroom": 39,
    "beige": 79,
    "birthday": 30,
    "black": 60,
    "blonde": 77,
    "blue and white": 55,
    "boy": 45,
    "brown": 25,
    "bus": 53,
    "car": 63,
    "cat": 11,
    "chair": 61,
    "chopsticks": 71,
    "crossing": 32,
    "crown": 51,
    "cup": 29,
    "dog": 50,
    "donut": 31,
    "double": 4,
    "doughnut": 82,
    "down": 33,
    "fashion": 65,
    "forest": 72,
    "full": 36,
    "girl": 1,
    "gray": 37,
    "green": 80,
    "hat": 17,
    "hawaii": 26,
    "human": 78,
    "ice cream": 76,
    "jeep": 16,
    "king": 52,
    "leather": 73,
    "little girl": 41,
    "many": 9,
    "no": 23,
    "not sure": 7,
    "out": 0,
    "outside": 86,
    "park": 13,
    "picnic table": 75,
    "pink": 42,
    "plastic": 28,
    "queen": 6,
    "red": 46,
    "red and blue": 19,
    "red and yellow": 18,
    "resting": 66,
    "right": 84,
    "shadow": 47,
    "shadows": 3,
    "shrimp": 49,
    "skateboard": 34,
    "sky": 24,
    "style": 38,
    "sun": 44,
    "suv": 14,
    "table": 27,
    "they aren't": 69,
    "tired": 10,
    "train": 64,
    "walking": 67,
    "wall": 56,
    "watching": 85,
    "wedding": 48,
    "white": 87,
    "white and black": 2,
    "white and blue": 74,
    "wine": 15,
    "wine tasting": 5,
    "woman": 12,
    "woods": 68,
    "yellow": 62,
    "yes": 83
  },
  "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": "4.37.2",
  "type_vocab_size": 2,
  "vocab_size": 30522
}