xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-25 13:13:42 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bug] Fix deepep low latency use nvlink by default (#27677)

Browse Source 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye 2025-10-28 19:53:12 -04:00 committed by GitHub
parent 94666612a9
commit d3ab240f39
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/envs.py
View File

@ -205,7 +205,7 @@ if TYPE_CHECKING:
VLLM_OBJECT_STORAGE_SHM_BUFFER_NAME: str = "VLLM_OBJECT_STORAGE_SHM_BUFFER"
VLLM_DEEPEP_BUFFER_SIZE_MB: int = 1024
VLLM_DEEPEP_HIGH_THROUGHPUT_FORCE_INTRA_NODE: bool = False
VLLM_DEEPEP_LOW_LATENCY_ALLOW_NVLINK: bool = False
VLLM_DEEPEP_LOW_LATENCY_ALLOW_NVLINK: bool = True
VLLM_DEEPEP_LOW_LATENCY_USE_MNNVL: bool = False
VLLM_DBO_COMM_SMS: int = 20
GPT_OSS_SYSTEM_TOOL_MCP_LABELS: list[str] = []
@ -1362,7 +1362,7 @@ environment_variables: dict[str, Callable[[], Any]] = {
# Allow DeepEP to use nvlink for internode_ll kernel, turn this on for
# better latency on GB200 like system
"VLLM_DEEPEP_LOW_LATENCY_ALLOW_NVLINK": lambda: bool(
int(os.getenv("VLLM_DEEPEP_LOW_LATENCY_ALLOW_NVLINK", "0"))
int(os.getenv("VLLM_DEEPEP_LOW_LATENCY_ALLOW_NVLINK", "1"))
),
# Allow DeepEP to use MNNVL (multi-node nvlink) for internode_ll kernel,
# turn this for better latency on GB200 like system