{ | |
"_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": "exit", | |
"1": "hat", | |
"2": "tired", | |
"3": "stand", | |
"4": "birthday", | |
"5": "curtain", | |
"6": "wine", | |
"7": "ball", | |
"8": "park", | |
"9": "jeep", | |
"10": "snowboard", | |
"11": "snow", | |
"12": "dirt", | |
"13": "lying down", | |
"14": "gray and black", | |
"15": "cat", | |
"16": "3", | |
"17": "soccer ball", | |
"18": "wall", | |
"19": "picnic table", | |
"20": "church", | |
"21": "shelter", | |
"22": "tan", | |
"23": "bricks", | |
"24": "white", | |
"25": "suv", | |
"26": "style", | |
"27": "ice cream", | |
"28": "down", | |
"29": "watching", | |
"30": "on street", | |
"31": "security", | |
"32": "yellow", | |
"33": "africa", | |
"34": "chair", | |
"35": "screen", | |
"36": "stripes", | |
"37": "tower", | |
"38": "5", | |
"39": "gray", | |
"40": "double", | |
"41": "smile", | |
"42": "name tag", | |
"43": "blue and white", | |
"44": "skier", | |
"45": "2000", | |
"46": "cage", | |
"47": "skiing", | |
"48": "yes", | |
"49": "woods", | |
"50": "black", | |
"51": "platform", | |
"52": "sidewalk", | |
"53": "calico", | |
"54": "necklace", | |
"55": "man", | |
"56": "lg", | |
"57": "ground", | |
"58": "queen", | |
"59": "small", | |
"60": "skateboarding", | |
"61": "rack", | |
"62": "clock tower", | |
"63": "snowboarding", | |
"64": "red and yellow", | |
"65": "desert", | |
"66": "0", | |
"67": "tabby", | |
"68": "2013", | |
"69": "canopy", | |
"70": "7:35", | |
"71": "tv", | |
"72": "net", | |
"73": "plastic", | |
"74": "out", | |
"75": "woman", | |
"76": "king", | |
"77": "nothing", | |
"78": "they aren't", | |
"79": "door", | |
"80": "in car", | |
"81": "protection", | |
"82": "bicycles", | |
"83": "green", | |
"84": "6", | |
"85": "2010", | |
"86": "arrow", | |
"87": "giraffe", | |
"88": "low", | |
"89": "orange", | |
"90": "large", | |
"91": "purple", | |
"92": "not there", | |
"93": "human", | |
"94": "window", | |
"95": "don't know", | |
"96": "many", | |
"97": "blonde", | |
"98": "hair", | |
"99": "talking", | |
"100": "doughnut", | |
"101": "crown", | |
"102": "big ben", | |
"103": "fence", | |
"104": "no", | |
"105": "boy", | |
"106": "person", | |
"107": "8", | |
"108": "happy", | |
"109": "sun", | |
"110": "cup", | |
"111": "clock", | |
"112": "lady", | |
"113": "train", | |
"114": "leather", | |
"115": "bicycle", | |
"116": "car", | |
"117": "shadows", | |
"118": "on road", | |
"119": "sky", | |
"120": "shade", | |
"121": "unknown", | |
"122": "curtains", | |
"123": "shrimp", | |
"124": "7:45", | |
"125": "solid", | |
"126": "little girl", | |
"127": "station", | |
"128": "trees", | |
"129": "fashion", | |
"130": "clear", | |
"131": "natural", | |
"132": "snowboarder", | |
"133": "hawaii", | |
"134": "backpack", | |
"135": "monitor", | |
"136": "9:35", | |
"137": "can't tell", | |
"138": "blue", | |
"139": "red and blue", | |
"140": "2", | |
"141": "windows", | |
"142": "girl", | |
"143": "full", | |
"144": "dog", | |
"145": "zoo", | |
"146": "women", | |
"147": "donut", | |
"148": "cross", | |
"149": "plain", | |
"150": "bedroom", | |
"151": "crossing", | |
"152": "8:35", | |
"153": "sleeping", | |
"154": "wedding", | |
"155": "10", | |
"156": "bike rack", | |
"157": "right", | |
"158": "white and blue", | |
"159": "wine tasting", | |
"160": "street", | |
"161": "cloudy", | |
"162": "7", | |
"163": "resting", | |
"164": "tent", | |
"165": "brick", | |
"166": "at table", | |
"167": "chopsticks", | |
"168": "photographer", | |
"169": "smiling", | |
"170": "bikes", | |
"171": "giraffes", | |
"172": "red", | |
"173": "not sure", | |
"174": "white and black", | |
"175": "1", | |
"176": "table", | |
"177": "shadow", | |
"178": "skateboard", | |
"179": "soccer", | |
"180": "bus", | |
"181": "black and white", | |
"182": "air", | |
"183": "talking on phone", | |
"184": "walking", | |
"185": "pink", | |
"186": "french", | |
"187": "camera", | |
"188": "forest", | |
"189": "lanyard", | |
"190": "beige", | |
"191": "neon", | |
"192": "beagle", | |
"193": "outside", | |
"194": "brown", | |
"195": "4", | |
"196": "roof", | |
"197": "plate", | |
"198": "laying down" | |
}, | |
"image_size": 384, | |
"initializer_range": 0.02, | |
"intermediate_size": 3072, | |
"label2id": { | |
"0": 66, | |
"1": 175, | |
"10": 155, | |
"2": 140, | |
"2000": 45, | |
"2010": 85, | |
"2013": 68, | |
"3": 16, | |
"4": 195, | |
"5": 38, | |
"6": 84, | |
"7": 162, | |
"7:35": 70, | |
"7:45": 124, | |
"8": 107, | |
"8:35": 152, | |
"9:35": 136, | |
"africa": 33, | |
"air": 182, | |
"arrow": 86, | |
"at table": 166, | |
"backpack": 134, | |
"ball": 7, | |
"beagle": 192, | |
"bedroom": 150, | |
"beige": 190, | |
"bicycle": 115, | |
"bicycles": 82, | |
"big ben": 102, | |
"bike rack": 156, | |
"bikes": 170, | |
"birthday": 4, | |
"black": 50, | |
"black and white": 181, | |
"blonde": 97, | |
"blue": 138, | |
"blue and white": 43, | |
"boy": 105, | |
"brick": 165, | |
"bricks": 23, | |
"brown": 194, | |
"bus": 180, | |
"cage": 46, | |
"calico": 53, | |
"camera": 187, | |
"can't tell": 137, | |
"canopy": 69, | |
"car": 116, | |
"cat": 15, | |
"chair": 34, | |
"chopsticks": 167, | |
"church": 20, | |
"clear": 130, | |
"clock": 111, | |
"clock tower": 62, | |
"cloudy": 161, | |
"cross": 148, | |
"crossing": 151, | |
"crown": 101, | |
"cup": 110, | |
"curtain": 5, | |
"curtains": 122, | |
"desert": 65, | |
"dirt": 12, | |
"dog": 144, | |
"don't know": 95, | |
"donut": 147, | |
"door": 79, | |
"double": 40, | |
"doughnut": 100, | |
"down": 28, | |
"exit": 0, | |
"fashion": 129, | |
"fence": 103, | |
"forest": 188, | |
"french": 186, | |
"full": 143, | |
"giraffe": 87, | |
"giraffes": 171, | |
"girl": 142, | |
"gray": 39, | |
"gray and black": 14, | |
"green": 83, | |
"ground": 57, | |
"hair": 98, | |
"happy": 108, | |
"hat": 1, | |
"hawaii": 133, | |
"human": 93, | |
"ice cream": 27, | |
"in car": 80, | |
"jeep": 9, | |
"king": 76, | |
"lady": 112, | |
"lanyard": 189, | |
"large": 90, | |
"laying down": 198, | |
"leather": 114, | |
"lg": 56, | |
"little girl": 126, | |
"low": 88, | |
"lying down": 13, | |
"man": 55, | |
"many": 96, | |
"monitor": 135, | |
"name tag": 42, | |
"natural": 131, | |
"necklace": 54, | |
"neon": 191, | |
"net": 72, | |
"no": 104, | |
"not sure": 173, | |
"not there": 92, | |
"nothing": 77, | |
"on road": 118, | |
"on street": 30, | |
"orange": 89, | |
"out": 74, | |
"outside": 193, | |
"park": 8, | |
"person": 106, | |
"photographer": 168, | |
"picnic table": 19, | |
"pink": 185, | |
"plain": 149, | |
"plastic": 73, | |
"plate": 197, | |
"platform": 51, | |
"protection": 81, | |
"purple": 91, | |
"queen": 58, | |
"rack": 61, | |
"red": 172, | |
"red and blue": 139, | |
"red and yellow": 64, | |
"resting": 163, | |
"right": 157, | |
"roof": 196, | |
"screen": 35, | |
"security": 31, | |
"shade": 120, | |
"shadow": 177, | |
"shadows": 117, | |
"shelter": 21, | |
"shrimp": 123, | |
"sidewalk": 52, | |
"skateboard": 178, | |
"skateboarding": 60, | |
"skier": 44, | |
"skiing": 47, | |
"sky": 119, | |
"sleeping": 153, | |
"small": 59, | |
"smile": 41, | |
"smiling": 169, | |
"snow": 11, | |
"snowboard": 10, | |
"snowboarder": 132, | |
"snowboarding": 63, | |
"soccer": 179, | |
"soccer ball": 17, | |
"solid": 125, | |
"stand": 3, | |
"station": 127, | |
"street": 160, | |
"stripes": 36, | |
"style": 26, | |
"sun": 109, | |
"suv": 25, | |
"tabby": 67, | |
"table": 176, | |
"talking": 99, | |
"talking on phone": 183, | |
"tan": 22, | |
"tent": 164, | |
"they aren't": 78, | |
"tired": 2, | |
"tower": 37, | |
"train": 113, | |
"trees": 128, | |
"tv": 71, | |
"unknown": 121, | |
"walking": 184, | |
"wall": 18, | |
"watching": 29, | |
"wedding": 154, | |
"white": 24, | |
"white and black": 174, | |
"white and blue": 158, | |
"window": 94, | |
"windows": 141, | |
"wine": 6, | |
"wine tasting": 159, | |
"woman": 75, | |
"women": 146, | |
"woods": 49, | |
"yellow": 32, | |
"yes": 48, | |
"zoo": 145 | |
}, | |
"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": "4.33.1", | |
"type_vocab_size": 2, | |
"vocab_size": 30522 | |
} | |