xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-25 05:55:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

updated

Browse Source 
Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com>
This commit is contained in:
rshaw@neuralmagic.com 2025-06-30 01:30:06 +00:00
parent ee2a4b0889
commit dedb1a5424
1 changed files with 7 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
View File

@ -330,7 +330,10 @@ class NixlConnectorWorker:
self.block_size = vllm_config.cache_config.block_size
# Agent.
self.nixl_wrapper = NixlWrapper(str(uuid.uuid4()), None)
self.nixl_wrapper = NixlWrapper(str(uuid.uuid4()),
None,
num_workers=None,
num_shared_workers=16)
# Map of engine_id -> {rank0: agent_name0, rank1: agent_name1..}.
self._remote_agents: dict[str, dict[int, str]] = defaultdict(dict)
@ -986,9 +989,9 @@ class NixlConnectorWorker:
# Begin async xfer.
start = time.perf_counter()
# for handle in handles:
# self.nixl_wrapper.transfer(handle)
self.nixl_wrapper.transfer_batched(handles)
for handle in handles:
self.nixl_wrapper.transfer(handle)
# self.nixl_wrapper.transfer_batched(handles)
end = time.perf_counter()
logger.info("======== LAUNCH TIME: %s ========", end - start)