xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-19 18:27:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

updated

Browse Source 
Signed-off-by: Robert Shaw <robshaw@redhat.com>
This commit is contained in:
Robert Shaw 2025-07-21 00:00:40 +00:00
parent e81c277e6e
commit 1dcd90065d
1 changed files with 2 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/v1/engine/core_client.py
View File

@ -431,11 +431,11 @@ class MPClient(EngineCoreClient):
dp_rank = parallel_config.data_parallel_rank
dp_local_size = parallel_config.data_parallel_size_local
offline_mode = parallel_config.data_parallel_rank_local is not None
# If Internal DPLB, the Client manages local+remote EngineCores.
# Otherwise, the Client just manages the local EngineCores.
local_engines_only = (parallel_config.data_parallel_hybrid_lb
or parallel_config.data_parallel_external_lb)
# If External DPLB, Client manages local EngineCores.
# If Internal DPLB, Client manages local+remote EngineCores.
num_ranks = dp_local_size if local_engines_only else dp_size
self.engine_ranks_managed = ([dp_rank] if offline_mode else range(
dp_rank, dp_rank + num_ranks))
@ -1020,7 +1020,6 @@ class DPLBAsyncMPClient(DPAsyncMPClient):
self, request: EngineCoreRequest) -> EngineIdentity:
# Engines are in rank order.
if (eng_index := request.data_parallel_rank) is None:
# logger.info(f"{self.lb_engines=} | {self.core_engines=}")
if not self.lb_engines:
return self.core_engine
# TODO use P2C alg for larger DP sizes