SGLang integration/maintenance tasks
Open
tracker
- Dominant language
- Python
- Stars
- 59
- Forks
- 32
- PR merge metrics
- No merged PRs in 30d
Description
This is a live issue tracker for all tasks related to Wave backend in SGLang.
Issue tracker:
- [ ] GPU memory fault (https://github.com/sgl-project/sglang/issues/10819)
- [ ] Unit test crash on MI300 (https://github.com/iree-org/wave/issues/97)
- [ ] Accuracy of decode kernel on MI350X and e2e run - https://github.com/iree-org/wave/issues/143
- [ ] Add sliding window support for extend and decode attention
- [ ] Use `sinks` and `xai_temperature_len` parameters in extend and decode kernels
Contributor guide
Assessment
This issue has not been assessed yet.