{
  "_name_or_path": "google/vit-base-patch16-224",
  "architectures": [
    "ViTForImageClassification"
  ],
  "attention_probs_dropout_prob": 0.0,
  "encoder_stride": 16,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 768,
  "id2label": {
    "0": "apple",
    "1": "ball",
    "10": "cereal_box",
    "11": "coffee_mug",
    "12": "comb",
    "13": "dry_battery",
    "14": "flashlight",
    "15": "food_bag",
    "16": "food_box",
    "17": "food_can",
    "18": "food_cup",
    "19": "food_jar",
    "2": "banana",
    "20": "garlic",
    "21": "glue_stick",
    "22": "greens",
    "23": "hand_towel",
    "24": "instant_noodles",
    "25": "keyboard",
    "26": "kleenex",
    "27": "lemon",
    "28": "lightbulb",
    "29": "lime",
    "3": "bell_pepper",
    "30": "marker",
    "31": "mushroom",
    "32": "notebook",
    "33": "onion",
    "34": "orange",
    "35": "peach",
    "36": "pear",
    "37": "pitcher",
    "38": "plate",
    "39": "pliers",
    "4": "binder",
    "40": "potato",
    "41": "rubber_eraser",
    "42": "scissors",
    "43": "shampoo",
    "44": "soda_can",
    "45": "sponge",
    "46": "stapler",
    "47": "tomato",
    "48": "toothbrush",
    "49": "toothpaste",
    "5": "bowl",
    "50": "water_bottle",
    "6": "calculator",
    "7": "camera",
    "8": "cap",
    "9": "cell_phone"
  },
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "apple": "0",
    "ball": "1",
    "banana": "2",
    "bell_pepper": "3",
    "binder": "4",
    "bowl": "5",
    "calculator": "6",
    "camera": "7",
    "cap": "8",
    "cell_phone": "9",
    "cereal_box": "10",
    "coffee_mug": "11",
    "comb": "12",
    "dry_battery": "13",
    "flashlight": "14",
    "food_bag": "15",
    "food_box": "16",
    "food_can": "17",
    "food_cup": "18",
    "food_jar": "19",
    "garlic": "20",
    "glue_stick": "21",
    "greens": "22",
    "hand_towel": "23",
    "instant_noodles": "24",
    "keyboard": "25",
    "kleenex": "26",
    "lemon": "27",
    "lightbulb": "28",
    "lime": "29",
    "marker": "30",
    "mushroom": "31",
    "notebook": "32",
    "onion": "33",
    "orange": "34",
    "peach": "35",
    "pear": "36",
    "pitcher": "37",
    "plate": "38",
    "pliers": "39",
    "potato": "40",
    "rubber_eraser": "41",
    "scissors": "42",
    "shampoo": "43",
    "soda_can": "44",
    "sponge": "45",
    "stapler": "46",
    "tomato": "47",
    "toothbrush": "48",
    "toothpaste": "49",
    "water_bottle": "50"
  },
  "layer_norm_eps": 1e-12,
  "model_type": "vit",
  "num_attention_heads": 12,
  "num_channels": 3,
  "num_hidden_layers": 12,
  "patch_size": 16,
  "qkv_bias": true,
  "transformers_version": "4.31.0"
}