{
  "architectures": [
    "ViltForQuestionAnswering"
  ],
  "attention_probs_dropout_prob": 0.0,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 768,
  "id2label": {
    "0": "walking",
    "1": "king",
    "2": "blonde",
    "3": "air",
    "4": "8",
    "5": "bedroom",
    "6": "ice cream",
    "7": "car",
    "8": "picnic table",
    "9": "white and blue",
    "10": "cup",
    "11": "crossing",
    "12": "human",
    "13": "0",
    "14": "resting",
    "15": "no",
    "16": "little girl",
    "17": "they aren't",
    "18": "woman",
    "19": "fashion",
    "20": "double",
    "21": "outside",
    "22": "3",
    "23": "train",
    "24": "sun",
    "25": "at table",
    "26": "hat",
    "27": "full",
    "28": "1",
    "29": "crown",
    "30": "woods",
    "31": "wine tasting",
    "32": "boy",
    "33": "dog",
    "34": "chopsticks",
    "35": "right",
    "36": "park",
    "37": "blue and white",
    "38": "forest",
    "39": "beige",
    "40": "cat",
    "41": "pink",
    "42": "skateboard",
    "43": "wall",
    "44": "plastic",
    "45": "not sure",
    "46": "2",
    "47": "red and blue",
    "48": "7",
    "49": "red and yellow",
    "50": "suv",
    "51": "yellow",
    "52": "gray",
    "53": "shadows",
    "54": "down",
    "55": "hawaii",
    "56": "girl",
    "57": "red",
    "58": "backpack",
    "59": "yes",
    "60": "shrimp",
    "61": "chair",
    "62": "watching",
    "63": "tired",
    "64": "jeep",
    "65": "10",
    "66": "wine",
    "67": "sky",
    "68": "5",
    "69": "leather",
    "70": "4",
    "71": "queen",
    "72": "bus",
    "73": "green",
    "74": "white and black",
    "75": "white",
    "76": "wedding",
    "77": "style",
    "78": "table",
    "79": "out",
    "80": "black",
    "81": "shadow",
    "82": "doughnut",
    "83": "donut",
    "84": "many",
    "85": "birthday",
    "86": "6",
    "87": "brown"
  },
  "image_size": 384,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "0": 13,
    "1": 28,
    "10": 65,
    "2": 46,
    "3": 22,
    "4": 70,
    "5": 68,
    "6": 86,
    "7": 48,
    "8": 4,
    "air": 3,
    "at table": 25,
    "backpack": 58,
    "bedroom": 5,
    "beige": 39,
    "birthday": 85,
    "black": 80,
    "blonde": 2,
    "blue and white": 37,
    "boy": 32,
    "brown": 87,
    "bus": 72,
    "car": 7,
    "cat": 40,
    "chair": 61,
    "chopsticks": 34,
    "crossing": 11,
    "crown": 29,
    "cup": 10,
    "dog": 33,
    "donut": 83,
    "double": 20,
    "doughnut": 82,
    "down": 54,
    "fashion": 19,
    "forest": 38,
    "full": 27,
    "girl": 56,
    "gray": 52,
    "green": 73,
    "hat": 26,
    "hawaii": 55,
    "human": 12,
    "ice cream": 6,
    "jeep": 64,
    "king": 1,
    "leather": 69,
    "little girl": 16,
    "many": 84,
    "no": 15,
    "not sure": 45,
    "out": 79,
    "outside": 21,
    "park": 36,
    "picnic table": 8,
    "pink": 41,
    "plastic": 44,
    "queen": 71,
    "red": 57,
    "red and blue": 47,
    "red and yellow": 49,
    "resting": 14,
    "right": 35,
    "shadow": 81,
    "shadows": 53,
    "shrimp": 60,
    "skateboard": 42,
    "sky": 67,
    "style": 77,
    "sun": 24,
    "suv": 50,
    "table": 78,
    "they aren't": 17,
    "tired": 63,
    "train": 23,
    "walking": 0,
    "wall": 43,
    "watching": 62,
    "wedding": 76,
    "white": 75,
    "white and black": 74,
    "white and blue": 9,
    "wine": 66,
    "wine tasting": 31,
    "woman": 18,
    "woods": 30,
    "yellow": 51,
    "yes": 59
  },
  "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": "4.52.4",
  "type_vocab_size": 2,
  "vocab_size": 30522
}