[BUGFIX] fix edge case for qwen3-next (#14209)

This commit is contained in:
Yi Zhang
2025-12-26 17:14:37 +08:00
committed by GitHub
parent 73c0c66ff8
commit ef92b4eb88
2 changed files with 9 additions and 7 deletions
+4 -1
View File
@@ -2583,7 +2583,10 @@ class Scheduler(
release_kv_cache(req, self.tree_cache)
# For mamba radix cache
if req.mamba_pool_idx is not None:
if (
req.mamba_pool_idx is not None
and self.disaggregation_mode != DisaggregationMode.DECODE
):
release_kv_cache(req, self.tree_cache, is_insert=False)
logger.debug(f"Abort queued request. {req.rid=}")