Update README.md
Browse files
README.md
CHANGED
|
@@ -20,7 +20,7 @@ license: apache-2.0
|
|
| 20 |
## Model Overview
|
| 21 |
- This model is finetuned with [VSA](https://arxiv.org/pdf/2505.13389), based on [Wan-AI/Wan2.1-T2V-14B-Diffusers](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B-Diffusers).
|
| 22 |
- It achieves up to 2.1x speed up on a single **H100** GPU.
|
| 23 |
-
- Our model is trained on **77×768×1280** resolution, but it supports generating videos with any resolution
|
| 24 |
- We set **VSA attention sparsity** to 0.9, and training runs for **1500 steps (~14 hours)**. You can tune this value from 0 to 0.9 to balance speed and performance for inference.
|
| 25 |
- Finetuning and inference scripts are available in the [FastVideo](https://github.com/hao-ai-lab/FastVideo) repository:
|
| 26 |
- [1 Node/GPU debugging finetuning script](https://github.com/hao-ai-lab/FastVideo/blob/main/scripts/finetune/finetune_v1_VSA.sh)
|
|
|
|
| 20 |
## Model Overview
|
| 21 |
- This model is finetuned with [VSA](https://arxiv.org/pdf/2505.13389), based on [Wan-AI/Wan2.1-T2V-14B-Diffusers](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B-Diffusers).
|
| 22 |
- It achieves up to 2.1x speed up on a single **H100** GPU.
|
| 23 |
+
- Our model is trained on **77×768×1280** resolution, but it supports generating videos with **any resolution**.(quality may degrade).
|
| 24 |
- We set **VSA attention sparsity** to 0.9, and training runs for **1500 steps (~14 hours)**. You can tune this value from 0 to 0.9 to balance speed and performance for inference.
|
| 25 |
- Finetuning and inference scripts are available in the [FastVideo](https://github.com/hao-ai-lab/FastVideo) repository:
|
| 26 |
- [1 Node/GPU debugging finetuning script](https://github.com/hao-ai-lab/FastVideo/blob/main/scripts/finetune/finetune_v1_VSA.sh)
|