{
  "_name_or_path": "phonghoccode/vilt-vqa-finetune-pytorch",
  "architectures": [
    "CustomViltForVQA"
  ],
  "attention_probs_dropout_prob": 0.0,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 768,
  "id2label": {
    "0": "donut",
    "1": "Orange",
    "2": "bottle",
    "3": "laptop",
    "4": "toilet",
    "5": "car",
    "6": "fork",
    "7": "bus",
    "8": "keyboard",
    "9": "0",
    "10": "Green",
    "11": "vase",
    "12": "bear",
    "13": "4",
    "14": "3",
    "15": "bird",
    "16": "Yellow",
    "17": "Grey",
    "18": "suitcase",
    "19": "skateboard",
    "20": "tv",
    "21": "Red",
    "22": "Behind",
    "23": "spoon",
    "24": "bicycle",
    "25": "7",
    "26": "remote",
    "27": "kite",
    "28": "orange",
    "29": "cow",
    "30": "Below",
    "31": "parking meter",
    "32": "right",
    "33": "bowl",
    "34": "sheep",
    "35": "handbag",
    "36": "potted plant",
    "37": "left",
    "38": "airplane",
    "39": "2",
    "40": "6",
    "41": "elephant",
    "42": "skis",
    "43": "bench",
    "44": "dog",
    "45": "truck",
    "46": "Left",
    "47": "umbrella",
    "48": "motorcycle",
    "49": "5",
    "50": "tennis racket",
    "51": "cake",
    "52": "Front",
    "53": "clock",
    "54": "teddy bear",
    "55": "hot dog",
    "56": "oven",
    "57": "toothbrush",
    "58": "Black",
    "59": "book",
    "60": "1",
    "61": "tie",
    "62": "couch",
    "63": "mouse",
    "64": "Brown",
    "65": "dining table",
    "66": "Pink",
    "67": "carrot",
    "68": "surfboard",
    "69": "pizza",
    "70": "bed",
    "71": "cell phone",
    "72": "broccoli",
    "73": "scissors",
    "74": "Purple",
    "75": "boat",
    "76": "Yes",
    "77": "apple",
    "78": "Blue",
    "79": "stop sign",
    "80": "8",
    "81": "frisbee",
    "82": "sports ball",
    "83": "9",
    "84": "fire hydrant",
    "85": "wine glass",
    "86": "sink",
    "87": "baseball glove",
    "88": "cat",
    "89": "train",
    "90": "banana",
    "91": "horse",
    "92": "above",
    "93": "White",
    "94": "traffic light",
    "95": "snowboard",
    "96": "No",
    "97": "baseball bat",
    "98": "person",
    "99": "refrigerator",
    "100": "zebra",
    "101": "chair",
    "102": "cup",
    "103": "giraffe",
    "104": "knife",
    "105": "Right",
    "106": "Above",
    "107": "sandwich"
  },
  "image_size": 384,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "0": 9,
    "1": 60,
    "2": 39,
    "3": 14,
    "4": 13,
    "5": 49,
    "6": 40,
    "7": 25,
    "8": 80,
    "9": 83,
    "Above": 106,
    "Behind": 22,
    "Below": 30,
    "Black": 58,
    "Blue": 78,
    "Brown": 64,
    "Front": 52,
    "Green": 10,
    "Grey": 17,
    "Left": 46,
    "No": 96,
    "Orange": 1,
    "Pink": 66,
    "Purple": 74,
    "Red": 21,
    "Right": 105,
    "White": 93,
    "Yellow": 16,
    "Yes": 76,
    "above": 92,
    "airplane": 38,
    "apple": 77,
    "banana": 90,
    "baseball bat": 97,
    "baseball glove": 87,
    "bear": 12,
    "bed": 70,
    "bench": 43,
    "bicycle": 24,
    "bird": 15,
    "boat": 75,
    "book": 59,
    "bottle": 2,
    "bowl": 33,
    "broccoli": 72,
    "bus": 7,
    "cake": 51,
    "car": 5,
    "carrot": 67,
    "cat": 88,
    "cell phone": 71,
    "chair": 101,
    "clock": 53,
    "couch": 62,
    "cow": 29,
    "cup": 102,
    "dining table": 65,
    "dog": 44,
    "donut": 0,
    "elephant": 41,
    "fire hydrant": 84,
    "fork": 6,
    "frisbee": 81,
    "giraffe": 103,
    "handbag": 35,
    "horse": 91,
    "hot dog": 55,
    "keyboard": 8,
    "kite": 27,
    "knife": 104,
    "laptop": 3,
    "left": 37,
    "motorcycle": 48,
    "mouse": 63,
    "orange": 28,
    "oven": 56,
    "parking meter": 31,
    "person": 98,
    "pizza": 69,
    "potted plant": 36,
    "refrigerator": 99,
    "remote": 26,
    "right": 32,
    "sandwich": 107,
    "scissors": 73,
    "sheep": 34,
    "sink": 86,
    "skateboard": 19,
    "skis": 42,
    "snowboard": 95,
    "spoon": 23,
    "sports ball": 82,
    "stop sign": 79,
    "suitcase": 18,
    "surfboard": 68,
    "teddy bear": 54,
    "tennis racket": 50,
    "tie": 61,
    "toilet": 4,
    "toothbrush": 57,
    "traffic light": 94,
    "train": 89,
    "truck": 45,
    "tv": 20,
    "umbrella": 47,
    "vase": 11,
    "wine glass": 85,
    "zebra": 100
  },
  "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": "4.35.2",
  "type_vocab_size": 2,
  "vocab_size": 30522
}