| license: apache-2.0 | |
| tags: | |
| - video-generation | |
| - streaming | |
| - WaveForcing | |
| - 1.3B | |
| # WaveForcing-1.3B | |
| Inference weights of the WaveForcing 1.3B text-to-video model, as native-Wan `model.safetensors`. Two variants: `5step/` (5 denoising steps) and `4step/` (4 denoising steps). | |
| Text encoder, tokenizer and VAE are not included; load them from [Wan2.1-T2V-1.3B](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B-Diffusers). | |
| ## Download | |
| ```bash | |
| hf download SJTU-DENG-Lab/WaveForcing-1.3B --local-dir ckpts/WaveForcing-1.3B | |
| ``` | |
| ## Run (WaveRT) | |
| 5-step: | |
| ```bash | |
| python -m wave_rt \ | |
| --model-path ckpts/Wan2.1-T2V-1.3B-Diffusers \ | |
| --gen-ckpt ckpts/WaveForcing-1.3B/5step/model.safetensors \ | |
| --rf-step 5 --wp-size 6 \ | |
| --denoising-step-list 1000,800,600,400,200 \ | |
| --vae-stages 2 --vae-partition time \ | |
| --kv-context causal --exchange-mode paged \ | |
| --attention-backend torch_sdpa | |
| ``` | |
| 4-step: point `--gen-ckpt` to `4step/model.safetensors` and use `--rf-step 4 --wp-size 5 --denoising-step-list 1000,750,500,250 --vae-stages 3`. | |
| Code: [SJTU-DENG-Lab/WaveForcing](https://github.com/SJTU-DENG-Lab/WaveForcing) | |