|
1 | 1 | """VideoPrism model preset configurations.""" |
2 | 2 |
|
3 | | -backbone_presets = {} |
| 3 | +backbone_presets = { |
| 4 | + "videoprism_public_v1_base": { |
| 5 | + "metadata": { |
| 6 | + "description": ( |
| 7 | + "114 million parameter, 12-layer ViT-B, 16-frame, 288x288 " |
| 8 | + "resolution, video-only encoder for " |
| 9 | + "spatiotemporal representation." |
| 10 | + ), |
| 11 | + "params": 114000000, |
| 12 | + "path": "video_prism", |
| 13 | + }, |
| 14 | + "kaggle_handle": "kaggle://keras/videoprism/keras/videoprism_public_v1_base/1", |
| 15 | + }, |
| 16 | + "videoprism_public_v1_large": { |
| 17 | + "metadata": { |
| 18 | + "description": ( |
| 19 | + "354 million parameter, 24-layer ViT-L, 16-frame, 288x288 " |
| 20 | + "resolution, video-only encoder for " |
| 21 | + "spatiotemporal representation." |
| 22 | + ), |
| 23 | + "params": 354000000, |
| 24 | + "path": "video_prism", |
| 25 | + }, |
| 26 | + "kaggle_handle": "kaggle://keras/videoprism/keras/videoprism_public_v1_large/1", |
| 27 | + }, |
| 28 | + "videoprism_lvt_public_v1_base": { |
| 29 | + "metadata": { |
| 30 | + "description": ( |
| 31 | + "248 million parameter, 12-layer ViT-B video encoder + text " |
| 32 | + "encoder, 16-frame, 288x288 resolution, for multimodal " |
| 33 | + "video-language tasks." |
| 34 | + ), |
| 35 | + "params": 248000000, |
| 36 | + "path": "video_prism", |
| 37 | + }, |
| 38 | + "kaggle_handle": "kaggle://keras/videoprism/keras/videoprism_lvt_public_v1_base/1", |
| 39 | + }, |
| 40 | + "videoprism_lvt_public_v1_large": { |
| 41 | + "metadata": { |
| 42 | + "description": ( |
| 43 | + "580 million parameter, 24-layer ViT-L video encoder + text " |
| 44 | + "encoder, 16-frame, 288x288 resolution, for multimodal " |
| 45 | + "video-language tasks." |
| 46 | + ), |
| 47 | + "params": 580000000, |
| 48 | + "path": "video_prism", |
| 49 | + }, |
| 50 | + "kaggle_handle": "kaggle://keras/videoprism/keras/videoprism_lvt_public_v1_large/1", |
| 51 | + }, |
| 52 | +} |
0 commit comments