Safetensors
FLUX.1-dev-fp8-flumina / requirements.txt
aredden's picture
remove torchao dependency, quantize entirely via linear
d45a331
raw
history blame
202 Bytes
git+https://github.com/aredden/torch-cublas-hgemm.git@master
einops
PyTurboJPEG
pydantic
fastapi
bitsandbytes
loguru
transformers
tokenizers
sentencepiece
click
accelerate
quanto
pydash
pybase64
uvicorn