Can we use some of the optimization functions on other models unrelated to SD and diffusion ?

chengzeyi / stable-fast

Best inference performance optimization framework for HuggingFace Diffusers on NVIDIA GPUs.

MIT License

1.05k stars 59 forks source link

Can we use some of the optimization functions on other models unrelated to SD and diffusion ? #120

Open oxysoft opened 5 months ago

oxysoft commented 5 months ago

I see many functions that seem general enough. If we have a depth estimation model in the artistic video pipeline for example, is it possible to introduce them to various optimizations and compiler features of this repo?

Cheers

chengzeyi commented 5 months ago

@oxysoft Yes, you can adopt some fine-tuned kernels and graph-capture mechanisms from stable-fast if they can help.