{ "_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": "tower", "1": "little girl", "2": "tan", "3": "natural", "4": "chopsticks", "5": "10", "6": "soccer", "7": "snowboarding", "8": "bedroom", "9": "neon", "10": "tv", "11": "8:35", "12": "picnic table", "13": "out", "14": "bikes", "15": "monitor", "16": "rack", "17": "photographer", "18": "cage", "19": "snow", "20": "green", "21": "boy", "22": "human", "23": "women", "24": "queen", "25": "suv", "26": "giraffe", "27": "blue", "28": "8", "29": "big ben", "30": "blue and white", "31": "wine tasting", "32": "person", "33": "4", "34": "2013", "35": "church", "36": "sun", "37": "hat", "38": "backpack", "39": "pink", "40": "bicycles", "41": "curtain", "42": "cloudy", "43": "donut", "44": "woods", "45": "down", "46": "yellow", "47": "king", "48": "bicycle", "49": "7:45", "50": "stand", "51": "0", "52": "shade", "53": "shrimp", "54": "birthday", "55": "snowboard", "56": "beige", "57": "security", "58": "tired", "59": "calico", "60": "2000", "61": "resting", "62": "cat", "63": "style", "64": "zoo", "65": "double", "66": "clock tower", "67": "black and white", "68": "bike rack", "69": "name tag", "70": "at table", "71": "leather", "72": "white", "73": "fence", "74": "plain", "75": "arrow", "76": "black", "77": "shelter", "78": "wall", "79": "wedding", "80": "not there", "81": "outside", "82": "windows", "83": "ice cream", "84": "woman", "85": "red and blue", "86": "train", "87": "5", "88": "car", "89": "white and black", "90": "orange", "91": "small", "92": "sidewalk", "93": "nothing", "94": "2010", "95": "plate", "96": "shadow", "97": "on street", "98": "soccer ball", "99": "protection", "100": "blonde", "101": "forest", "102": "red and yellow", "103": "net", "104": "white and blue", "105": "dog", "106": "necklace", "107": "happy", "108": "talking on phone", "109": "walking", "110": "7:35", "111": "clock", "112": "skier", "113": "africa", "114": "exit", "115": "lanyard", "116": "skiing", "117": "lady", "118": "french", "119": "air", "120": "9:35", "121": "giraffes", "122": "in car", "123": "2", "124": "camera", "125": "they aren't", "126": "curtains", "127": "on road", "128": "bus", "129": "wine", "130": "lg", "131": "purple", "132": "man", "133": "brown", "134": "6", "135": "yes", "136": "hawaii", "137": "trees", "138": "shadows", "139": "unknown", "140": "large", "141": "1", "142": "talking", "143": "brick", "144": "jeep", "145": "3", "146": "dirt", "147": "cross", "148": "hair", "149": "park", "150": "smiling", "151": "skateboarding", "152": "not sure", "153": "canopy", "154": "clear", "155": "tabby", "156": "table", "157": "crown", "158": "window", "159": "tent", "160": "don't know", "161": "no", "162": "right", "163": "7", "164": "red", "165": "can't tell", "166": "full", "167": "plastic", "168": "doughnut", "169": "watching", "170": "screen", "171": "roof", "172": "crossing", "173": "low", "174": "girl", "175": "snowboarder", "176": "bricks", "177": "many", "178": "solid", "179": "stripes", "180": "ball", "181": "ground", "182": "street", "183": "smile", "184": "skateboard", "185": "sleeping", "186": "cup", "187": "station", "188": "sky", "189": "chair", "190": "desert", "191": "door", "192": "laying down", "193": "beagle", "194": "gray and black", "195": "platform", "196": "gray", "197": "fashion", "198": "lying down" }, "image_size": 384, "initializer_range": 0.02, "intermediate_size": 3072, "label2id": { "0": 51, "1": 141, "10": 5, "2": 123, "2000": 60, "2010": 94, "2013": 34, "3": 145, "4": 33, "5": 87, "6": 134, "7": 163, "7:35": 110, "7:45": 49, "8": 28, "8:35": 11, "9:35": 120, "africa": 113, "air": 119, "arrow": 75, "at table": 70, "backpack": 38, "ball": 180, "beagle": 193, "bedroom": 8, "beige": 56, "bicycle": 48, "bicycles": 40, "big ben": 29, "bike rack": 68, "bikes": 14, "birthday": 54, "black": 76, "black and white": 67, "blonde": 100, "blue": 27, "blue and white": 30, "boy": 21, "brick": 143, "bricks": 176, "brown": 133, "bus": 128, "cage": 18, "calico": 59, "camera": 124, "can't tell": 165, "canopy": 153, "car": 88, "cat": 62, "chair": 189, "chopsticks": 4, "church": 35, "clear": 154, "clock": 111, "clock tower": 66, "cloudy": 42, "cross": 147, "crossing": 172, "crown": 157, "cup": 186, "curtain": 41, "curtains": 126, "desert": 190, "dirt": 146, "dog": 105, "don't know": 160, "donut": 43, "door": 191, "double": 65, "doughnut": 168, "down": 45, "exit": 114, "fashion": 197, "fence": 73, "forest": 101, "french": 118, "full": 166, "giraffe": 26, "giraffes": 121, "girl": 174, "gray": 196, "gray and black": 194, "green": 20, "ground": 181, "hair": 148, "happy": 107, "hat": 37, "hawaii": 136, "human": 22, "ice cream": 83, "in car": 122, "jeep": 144, "king": 47, "lady": 117, "lanyard": 115, "large": 140, "laying down": 192, "leather": 71, "lg": 130, "little girl": 1, "low": 173, "lying down": 198, "man": 132, "many": 177, "monitor": 15, "name tag": 69, "natural": 3, "necklace": 106, "neon": 9, "net": 103, "no": 161, "not sure": 152, "not there": 80, "nothing": 93, "on road": 127, "on street": 97, "orange": 90, "out": 13, "outside": 81, "park": 149, "person": 32, "photographer": 17, "picnic table": 12, "pink": 39, "plain": 74, "plastic": 167, "plate": 95, "platform": 195, "protection": 99, "purple": 131, "queen": 24, "rack": 16, "red": 164, "red and blue": 85, "red and yellow": 102, "resting": 61, "right": 162, "roof": 171, "screen": 170, "security": 57, "shade": 52, "shadow": 96, "shadows": 138, "shelter": 77, "shrimp": 53, "sidewalk": 92, "skateboard": 184, "skateboarding": 151, "skier": 112, "skiing": 116, "sky": 188, "sleeping": 185, "small": 91, "smile": 183, "smiling": 150, "snow": 19, "snowboard": 55, "snowboarder": 175, "snowboarding": 7, "soccer": 6, "soccer ball": 98, "solid": 178, "stand": 50, "station": 187, "street": 182, "stripes": 179, "style": 63, "sun": 36, "suv": 25, "tabby": 155, "table": 156, "talking": 142, "talking on phone": 108, "tan": 2, "tent": 159, "they aren't": 125, "tired": 58, "tower": 0, "train": 86, "trees": 137, "tv": 10, "unknown": 139, "walking": 109, "wall": 78, "watching": 169, "wedding": 79, "white": 72, "white and black": 89, "white and blue": 104, "window": 158, "windows": 82, "wine": 129, "wine tasting": 31, "woman": 84, "women": 23, "woods": 44, "yellow": 46, "yes": 135, "zoo": 64 }, "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": "4.33.2", "type_vocab_size": 2, "vocab_size": 30522 }