config.json
920 B · 37 lines · json Raw
1 {
2 "_name_or_path": "./",
3 "model_type": "VideoMAEv2_Base",
4 "architectures": [
5 "VideoMAEv2_Base"
6 ],
7 "auto_map": {
8 "AutoModel": "modeling_videomaev2.VideoMAEv2",
9 "AutoConfig": "modeling_config.VideoMAEv2Config"
10 },
11 "model_config":{
12 "img_size": 224,
13 "patch_size": 16,
14 "in_chans": 3,
15 "num_classes": 0,
16 "embed_dim": 1280,
17 "depth": 32,
18 "num_heads": 16,
19 "mlp_ratio": 4,
20 "qkv_bias": true,
21 "qk_scale": null,
22 "drop_rate": 0.0,
23 "attn_drop_rate": 0.0,
24 "drop_path_rate": 0.0,
25 "norm_layer": "nn.LayerNorm",
26 "layer_norm_eps": 1e-6,
27 "init_values": 0.0,
28 "use_learnable_pos_emb": false,
29 "tubelet_size": 2,
30 "use_mean_pooling": true,
31 "with_cp": false,
32 "num_frames": 16,
33 "cos_attn": false
34 },
35 "transformers_version": "4.38.0",
36 "use_cache": true
37 }