[ATOM SGLang] SGL plugin Attention Refractory - #863

Merged

zhuyuhua-v merged 17 commits into

zejunchen-zejun:zhiwei/attn_refrac_integrated

Jun 3, 2026

Commits on Jun 2, 2026

add work log

ZhiweiYan-96
authored and
zhuyuhua-v
committed
[ATOM-SGL][Attn refrac] Separate model-specific MLA from SGL full attention backend

ZhiweiYan-96
authored and
zhuyuhua-v
committed
remove work log

ZhiweiYan-96
authored and
zhuyuhua-v
committed
[ATOM-SGL][Attn refrac] Route DeepSeek MLA through an SGLang wrapper

ZhiweiYan-96
authored and
zhuyuhua-v
committed
[ATOM SGL] runtime extraction

ZhiweiYan-96
authored and
zhuyuhua-v
committed
[ATOM-SGL][Runtime] Introduce model adapter specs

authored and
zhuyuhua-v
committed
[ATOM-SGL][Runtime] Keep custom wrappers out of generated entries

authored and
zhuyuhua-v
committed
[ATOM-SGL][Attn refrac] Split full attention backend helpers

authored and
zhuyuhua-v
committed
[ATOM-SGL][Attn refrac] Format refactored attention files

authored and
zhuyuhua-v
committed
[ATOM-SGL][Attn refrac] Fix ruff findings in refactored attention code

authored and
zhuyuhua-v
committed
[ATOM-SGL][Attn refrac] Avoid DeepSeek MLA wrapper module cycle

authored and
zhuyuhua-v
committed
fix rebase issue

ZhiweiYan-96
authored and
zhuyuhua-v
committed
precheckin

ZhiweiYan-96
authored and
zhuyuhua-v
committed
prepare for sglang only

ZhiweiYan-96
authored and
zhuyuhua-v
committed
import error meet in qwen3.5

ZhiweiYan-96
authored and
zhuyuhua-v
committed
qwen3.5 acc fix

ZhiweiYan-96
authored and
zhuyuhua-v
committed
[Fix] Limit static FP4 linear kv_b_proj post-processing

qichu-yun
authored and
zhuyuhua-v
committed