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 6, 15:45 UTC
[Bugfix][MultiConnector] Fix MultiConnector for SupportsHMA sub-connectors (#36549)
zhanqiuhu committed
Mar 16, 2026 at 16:51 UTC
2dccb38f73fa79bc629b8b215b8066e61ce4a211
1 file changed
+8
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+8
@@ -415,6 +415,14 @@ class NixlConnector(KVConnectorBase_V1, SupportsHMA):
415
assert self.connector_scheduler is not None
416
return self.connector_scheduler.build_connector_meta(scheduler_output)
417
418
+ def request_finished(
419
+ self,
420
+ request: "Request",
421
+ block_ids: list[int],
422
+ ) -> tuple[bool, dict[str, Any] | None]:
423
+ assert self.connector_scheduler is not None
424
+ return self.connector_scheduler.request_finished(request, (block_ids,))
425
+
426
def request_finished_all_groups(
427
self,
428
request: "Request",