Upload component_registry.json
Browse files- component_registry.json +56 -0
component_registry.json
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"components": [
|
| 3 |
+
{
|
| 4 |
+
"name": "video_backbone",
|
| 5 |
+
"local_path": "video-gen",
|
| 6 |
+
"hf_repo": "Wan-AI/Wan2.2-TI2V-5B",
|
| 7 |
+
"parameters": 5000000000
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"name": "vlm_backbone",
|
| 11 |
+
"local_path": "captioner",
|
| 12 |
+
"hf_repo": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
|
| 13 |
+
"parameters": 500000000
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"name": "audio_backbone",
|
| 17 |
+
"local_path": "audio-gen",
|
| 18 |
+
"hf_repo": "suno/bark-small",
|
| 19 |
+
"parameters": 420000000
|
| 20 |
+
}
|
| 21 |
+
],
|
| 22 |
+
"adapters": [
|
| 23 |
+
{
|
| 24 |
+
"name": "video_to_unified",
|
| 25 |
+
"type": "ModalityBridge"
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"name": "vlm_to_unified",
|
| 29 |
+
"type": "ModalityBridge"
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"name": "audio_to_unified",
|
| 33 |
+
"type": "ModalityBridge"
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"name": "unified_to_video",
|
| 37 |
+
"type": "ModalityBridge"
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"name": "unified_to_vlm",
|
| 41 |
+
"type": "ModalityBridge"
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"name": "unified_to_audio",
|
| 45 |
+
"type": "ModalityBridge"
|
| 46 |
+
}
|
| 47 |
+
],
|
| 48 |
+
"synchronization": {
|
| 49 |
+
"type": "AudioVideoTemporalSynchronizer",
|
| 50 |
+
"mechanisms": [
|
| 51 |
+
"cross_attention",
|
| 52 |
+
"temporal_gru",
|
| 53 |
+
"sync_map_generator"
|
| 54 |
+
]
|
| 55 |
+
}
|
| 56 |
+
}
|