{ "_name_or_path": "google/vivit-b-16x2-kinetics400", "architectures": [ "VivitForVideoClassification" ], "attention_probs_dropout_prob": 0.0, "hidden_act": "gelu_fast", "hidden_dropout_prob": 0.0, "hidden_size": 768, "id2label": { "0": "ApplyEyeMakeup", "1": "ApplyLipstick", "2": "BabyCrawling", "3": "BlowDryHair", "4": "BlowingCandles", "5": "BrushingTeeth", "6": "CuttingInKitchen", "7": "Haircut", "8": "HeadMassage", "9": "JugglingBalls", "10": "Knitting", "11": "MoppingFloor", "12": "PizzaTossing", "13": "PlayingCello", "14": "PlayingFlute", "15": "PlayingGuitar", "16": "PlayingPiano", "17": "PlayingSitar", "18": "PlayingTabla", "19": "PlayingViolin", "20": "ShavingBeard", "21": "Typing", "22": "WallPushups", "23": "WritingOnBoard", "24": "YoYo" }, "image_size": 224, "initializer_range": 0.02, "intermediate_size": 3072, "label2id": { "ApplyEyeMakeup": 0, "ApplyLipstick": 1, "BabyCrawling": 2, "BlowDryHair": 3, "BlowingCandles": 4, "BrushingTeeth": 5, "CuttingInKitchen": 6, "Haircut": 7, "HeadMassage": 8, "JugglingBalls": 9, "Knitting": 10, "MoppingFloor": 11, "PizzaTossing": 12, "PlayingCello": 13, "PlayingFlute": 14, "PlayingGuitar": 15, "PlayingPiano": 16, "PlayingSitar": 17, "PlayingTabla": 18, "PlayingViolin": 19, "ShavingBeard": 20, "Typing": 21, "WallPushups": 22, "WritingOnBoard": 23, "YoYo": 24 }, "layer_norm_eps": 1e-06, "model_type": "vivit", "num_attention_heads": 12, "num_channels": 3, "num_frames": 32, "num_hidden_layers": 12, "qkv_bias": true, "torch_dtype": "float32", "transformers_version": "4.47.1", "tubelet_size": [ 2, 16, 16 ], "video_size": [ 32, 224, 224 ] }