What's Changed
- rm unused lmdeploy patch by @jinminxi104 in #310
- fix prefix caching by @yao-fengchen in #307
- fix PA param list by @jinminxi104 in #314
- [Ascend] support qwen3next by @wanfengcxz in #304
- remove unused import by @jinminxi104 in #317
- fix ci by @jinminxi104 in #318
- add mix_prefill test case by @yao-fengchen in #320
- fix re-capture in RL by @jinminxi104 in #319
- optimize rope and remove unused l2norm_fwd by @wanfengcxz in #316
- fix state dtype and refactor gdn kernel by @wanfengcxz in #321
- Bump028 by @jinminxi104 in #323
Full Changelog: v0.2.7...v0.2.8
Contributors
jinminxi104, wanfengcxz, and yao-fengchen