A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 07:15 UTC
[Bugfix][P/D] Fix Prefix Cache Bug (#18411)
Signed-off-by: nicklucche <nlucches@redhat.com> Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Nicolò Lucchesi committed
Jun 4, 2025 at 01:53 UTC
5d96533e2235c37e64ef381fafa244db197b25dc
1 file changed
+2
-1
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+2
-1
@@ -739,7 +739,8 @@ class NixlConnectorWorker:
739
# just notify P worker that we have the blocks we need.
740
num_local_blocks = len(local_block_ids)
741
if num_local_blocks == 0:
742
- self.nixl_wrapper.send_notif(dst_engine_id,
742
+ agent_name = self._remote_agents[dst_engine_id]
743
+ self.nixl_wrapper.send_notif(agent_name,
744
notif_msg=request_id.encode("utf-8"))
745
return
746