[BUGFIX] fix edge case for qwen3-next (#14209)
This commit is contained in:
@@ -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=}")
|
||||
|
||||
|
||||
Reference in New Issue
Block a user