Instructions to use Comfy-Org/Qwen-Image-Edit_ComfyUI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusion Single File
How to use Comfy-Org/Qwen-Image-Edit_ComfyUI 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
could you please provide a fp8 e5m2 for all those with a 3090 or lower so we can torch compile?
could you please provide a fp8 e5m2 for all those with a 3090 or lower so we can torch compile?
You can use fp8 e4 enabled version of Triton here: https://github.com/Comfy-Org/wheels/actions/runs/17170280663
Torch compile works great with 30 series thanks to that
:o that's great to know, thanks!
One question though. If I use that. will all of my e5m2 files still compile as well?
@DarkJanissary
So, I'm using that and it works great. But do you know if they'll continue to make fp8 builds like that for new releases? I couldn't find much information on it.
Edit:
https://github.com/woct0rdho/triton-windows/releases/tag/v3.5.0-windows.post21
It's now merged into the main windows wheel release!
solved!