fpx_weight_only quantization

sayakpaul / diffusers-torchao

End-to-end recipes for optimizing diffusion models with torchao and diffusers (inference and FP8 training).

Apache License 2.0

274 stars 8 forks source link

Closed a-r-r-o-w closed 2 months ago

a-r-r-o-w commented 2 months ago

The benchmarks are currently running. Will update here when complete

sayakpaul commented 2 months ago

Would be cool to propagate the changes to benchmark_image.py once the benchmarks are done.

a-r-r-o-w commented 2 months ago

@sayakpaul Let's merge this and you can update the image benchmark numbers in another PR?

Also, currently fpx + compile seems to be failing somewhere within the torchao library. I will look into it soon.