[docs]: document the FastH3 VSA kernel requirement
Browse filesState prominently that this checkpoint requires FastVideo VSA-H3 and the Video Sparse Attention kernel.
README.md
CHANGED
|
@@ -29,6 +29,10 @@ synchronized video and audio with four transformer forwards and uses VSA-H3 at
|
|
| 29 |
See the complete [FastH3 4-Step Preview v1
|
| 30 |
collection](https://huggingface.co/collections/FastVideo/fastvideo-fasth3).
|
| 31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
## Checkpoint identity
|
| 33 |
|
| 34 |
- Training run: `20260823T132231Z-v10p5-datafree-sp4`
|
|
|
|
| 29 |
See the complete [FastH3 4-Step Preview v1
|
| 30 |
collection](https://huggingface.co/collections/FastVideo/fastvideo-fasth3).
|
| 31 |
|
| 32 |
+
> **VSA requirement:** This checkpoint must be run with FastVideo's Video
|
| 33 |
+
> Sparse Attention (VSA-H3) backend and kernel enabled at 90% sparsity with
|
| 34 |
+
> 64-token tiles. Dense attention is not a supported substitute.
|
| 35 |
+
|
| 36 |
## Checkpoint identity
|
| 37 |
|
| 38 |
- Training run: `20260823T132231Z-v10p5-datafree-sp4`
|