Core kernels to implement for Wave
Open
- Dominant language
- Python
- Stars
- 59
- Forks
- 32
- PR merge metrics
- No merged PRs in 30d
Description
Functionality
- [ ] Int8 Group-quantized linear / Dequant-GEMM
- [ ] Int4 Group-quantized linear / Dequant-GEMM
- [ ] MoE align block
- [ ] MLA ([link](https://github.com/ROCm/aiter/blob/main/aiter/mla.py))
Need polishing
- [ ] Flash attention backwards - Need to improve performance
- [ ] Quant Linear - Need to improve accuracy
Contributor guide
Assessment
This issue has not been assessed yet.