EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Qwen3-Next support (#10233)

Browse Source 
Co-authored-by: cao1zhg <114661107+cao1zhg@users.noreply.github.com>
Co-authored-by: ispobock <ispobaoke@gmail.com>
Co-authored-by: Binyao Jiang <byjiang1996@gmail.com>
Co-authored-by: hebiao064 <hebiaobuaa@gmail.com>
Co-authored-by: Lifu Huang <lifu.hlf@gmail.com>
Co-authored-by: qingquansong <ustcsqq@gmail.com>
Co-authored-by: Yaoyao Ding <dingyaoyao.cs@gmail.com>
Co-authored-by: Ke Bao <ISPObaoke@163.com>
Co-authored-by: Minglei Zhu <mingleizhu1122@gmail.com>
This commit is contained in:
Yi Zhang
2025-09-11 19:11:49 +08:00
committed by GitHub
parent bfe01a5eef
commit 30c6e1f569
19 changed files with 3224 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
python/sglang/srt/managers/scheduler.py
View File

@@ -1540,7 +1540,12 @@ class Scheduler(
chunked_req_to_exclude.add(self.chunked_req)
self.tree_cache.cache_unfinished_req(self.chunked_req, chunked=True)
# chunked request keeps its rid but will get a new req_pool_idx
self.req_to_token_pool.free(self.chunked_req.req_pool_idx)
if self.tp_worker.worker.model_runner.is_hybrid_gdn:
self.req_to_token_pool.free(
self.chunked_req.req_pool_idx, free_mamba_cache=False
)
else:
self.req_to_token_pool.free(self.chunked_req.req_pool_idx)
if self.last_batch and self.last_batch.forward_mode.is_extend():
if self.last_batch.chunked_req is not None:
# In the context pipeline parallelism, after the last chunk, the current microbatch still track outdated chunked_req.