config.json
1.7 KB · 78 lines · json Raw
1 {
2 "_name_or_path": "MCG-NJU/videomae-large",
3 "architectures": [
4 "VideoMAEForVideoClassification"
5 ],
6 "attention_probs_dropout_prob": 0.0,
7 "decoder_hidden_size": 512,
8 "decoder_intermediate_size": 2048,
9 "decoder_num_attention_heads": 8,
10 "decoder_num_hidden_layers": 12,
11 "hidden_act": "gelu",
12 "hidden_dropout_prob": 0.0,
13 "hidden_size": 1024,
14 "id2label": {
15 "0": "arc_left",
16 "1": "arc_right",
17 "2": "dolly_in",
18 "3": "dolly_out",
19 "4": "pan_left",
20 "5": "pan_right",
21 "6": "pedestal_down",
22 "7": "pedestal_up",
23 "8": "pov",
24 "9": "roll_left",
25 "10": "roll_right",
26 "11": "shake",
27 "12": "static",
28 "13": "tilt_down",
29 "14": "tilt_up",
30 "15": "track",
31 "16": "truck_left",
32 "17": "truck_right",
33 "18": "undefined",
34 "19": "zoom_in",
35 "20": "zoom_out"
36 },
37 "image_size": 224,
38 "initializer_range": 0.02,
39 "intermediate_size": 4096,
40 "label2id": {
41 "arc_left": 0,
42 "arc_right": 1,
43 "dolly_in": 2,
44 "dolly_out": 3,
45 "pan_left": 4,
46 "pan_right": 5,
47 "pedestal_down": 6,
48 "pedestal_up": 7,
49 "pov": 8,
50 "roll_left": 9,
51 "roll_right": 10,
52 "shake": 11,
53 "static": 12,
54 "tilt_down": 13,
55 "tilt_up": 14,
56 "track": 15,
57 "truck_left": 16,
58 "truck_right": 17,
59 "undefined": 18,
60 "zoom_in": 19,
61 "zoom_out": 20
62 },
63 "layer_norm_eps": 1e-12,
64 "model_type": "videomae",
65 "norm_pix_loss": true,
66 "num_attention_heads": 16,
67 "num_channels": 3,
68 "num_frames": 16,
69 "num_hidden_layers": 24,
70 "patch_size": 16,
71 "problem_type": "multi_label_classification",
72 "qkv_bias": true,
73 "torch_dtype": "float32",
74 "transformers_version": "4.48.3",
75 "tubelet_size": 2,
76 "use_mean_pooling": false
77 }
78