Instructions to use lightx2v/Wan2.1-T2V-14B-StepDistill-CfgDistill with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use lightx2v/Wan2.1-T2V-14B-StepDistill-CfgDistill with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("lightx2v/Wan2.1-T2V-14B-StepDistill-CfgDistill", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Diffusion Single File
How to use lightx2v/Wan2.1-T2V-14B-StepDistill-CfgDistill with Diffusion Single File:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
which sampler is recommended for this model? I've tried several common ones, but none of them work well.
looking forward to your suggestion
Hi, we use a scheduler with fixed denoising time steps. For the four-step model, we use fixed time steps of 1000, 750, 500, and 250. See lightx2v-run_wan_t2v_distill.sh.
Hi, we use a scheduler with fixed denoising time steps. For the four-step model, we use fixed time steps of 1000, 750, 500, and 250. See lightx2v-run_wan_t2v_distill.sh.
This doesn't answer his question. We all know that LCM is the recommended Sampler, but what is the recommended Scheduler? Normal/Simple/Beta/Sgm_Uniform? This is the question.
Based on WanStepDistillScheduler it seems like you want linear. So for 4 steps (timesteps [1000, 750, 500, 250]) you'd want sigmas [1.0, 0.75, 0.5, 0.25, 0.0].