aredden / flux-fp8-api

Flux diffusion model implementation using quantized fp8 matmul & remaining layers use faster half precision accumulate, which is ~2x faster on consumer devices.
Apache License 2.0
202 stars 21 forks source link

WHL files for torch-cublas-hgemm.git and ao? #7

Closed SoftologyPro closed 2 months ago

SoftologyPro commented 2 months ago

Does anyone have WHL files for these to install on Windows? I am having all sorts of problems trying to build WHLs myself and cannot get the pip installs of these 2 working.

pip install git+https://github.com/aredden/torch-cublas-hgemm.git@master pip install git+https://github.com/pytorch/ao.git@main

Thanks.

SoftologyPro commented 2 months ago

For anyone else, do not run that exe inside that zip... https://www.virustotal.com/gui/file/081db758b854442bdfd28da76e67632d97e544f83334f38ca10854c2ef2810bb

aredden commented 2 months ago

Ah- no there aren't any WHL's, it's just this repo, sorry 😢