Plans for a Triton implementation?

Open arnavdantuluri opened this issue 2 years ago • 2 comments

Thank you guys for open sourcing this amazing work, I was curious though if there are any plans for a Triton implementation for a higher level implementation. I would like to experiment with this project in tandem with a library I have been working on to accelerate diffusion models but I am not entirely familiar with CUDA yet.

Looking forward to your response 🙂

Nov 27 '23 04:11 arnavdantuluri

We may try if we have free time, but it's not in the pipeline at the moment. We welcome community contributions!

What changes are you interested in making that Triton would be helpful for?

Nov 27 '23 21:11 DanFu09

Mainly looking for an implementation I can easily play around with, hopefully stuff like bias and activation fusion, extension to 2D, etc. Is there a reference pytorch implementation anywhere I can look at?

Nov 30 '23 00:11 arnavdantuluri