FP8 seems not working on RDNA4
Open
Feature
- Dominant language
- Python
- Stars
- 133k
- Forks
- 15.7k
- Avg merge
- 1d 7h
- Merged PRs (30d)
- 158
Description
### Feature Idea
Trying latest ROCm7 with Pytorch for ROCm 7.0 but still not using FP8 even when trying to force with:
--supports-fp8-compute
Returning this when executing:
"model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16"
GPU is a RX 9070 RDNA4
### Existing Solutions
_No response_
### Other
_No response_
Contributor guide
Assessment
This issue has not been assessed yet.