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, 09:45 UTC
[Bugfix][Failing Test] Fix nixl connector test when promt size < block size (#18429)
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
wwl2755 committed
May 21, 2025 at 00:41 UTC
d06dd72ba9d87794f545b08907426fe083965ade
1 file changed
+9
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+9
@@ -259,6 +259,15 @@ class NixlConnectorScheduler:
259
# Loop through scheduled reqs and convert to ReqMeta.
260
for req_id, (req, block_ids) in self._reqs_need_recv.items():
261
assert req.kv_transfer_params is not None
262
+ # For the case where there are no remote blocks to pull
263
+ # (block_ids is empty), we don't need to schedule
264
+ # an async read on the worker side.
265
+ if not block_ids:
266
+ logger.debug(
267
+ "Skipping adding request %s to NixlConnectorMetadata, "
268
+ "as there are no remote blocks to pull", req_id)
269
+ continue
270
+
271
meta.add_new_req(
272
request_id=req_id,
273
local_block_ids=block_ids,