Comfy-Org / Comfy-Org/comfy-kitchen
FP8 E4M3 quantized models inference is slower than same model quantized in MXFP8
Open
- Dominant language
- Python
- Stars
- 220
- Forks
- 91
- Avg merge
- 1d 7h
- Merged PRs (30d)
- 12
Description
It's quite weird. I think mxfp8 inference is a bit more complex than the legacy fp8e4m3. So does it mean we still have some potential in legacy fp8 inference?
Contributor guide
Assessment
This issue has not been assessed yet.