[Feature] Add --cpu-distributed-timeout-seconds CLI Option for CPU Process Group Timeout (#42968)

Signed-off-by: fangyuchu <fangyuchu@qq.com>
Signed-off-by: zWaNg3 <389750525@qq.com>
Co-authored-by: zWaNg3 <389750525@qq.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
fangyuchu
2026-05-21 15:42:07 -07:00
committed by GitHub
co-authored by zWaNg3 mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 0f66623b0d
commit e26e1f0928
4 changed files with 33 additions and 1 deletions
+14
View File
@@ -491,6 +491,16 @@ def get_cached_tcp_store_client(host: str, port: int) -> TCPStore:
return TCPStore(host, port, is_master=False, wait_for_workers=False)
def get_cpu_distributed_timeout_or_none() -> timedelta | None:
from vllm.config import get_current_vllm_config_or_none
vllm_config = get_current_vllm_config_or_none()
if vllm_config is None:
return None
timeout_seconds = vllm_config.parallel_config.cpu_distributed_timeout_seconds
return timedelta(seconds=timeout_seconds) if timeout_seconds is not None else None
def init_gloo_process_group(
prefix_store: PrefixStore,
group_rank: int,
@@ -570,6 +580,10 @@ def stateless_init_torch_distributed_process_group(
init_method = get_tcp_uri(host, port)
backend = Backend(backend) # it is basically string
timeout = _get_default_timeout(backend)
if backend == "gloo":
gloo_timeout = get_cpu_distributed_timeout_or_none()
if gloo_timeout is not None:
timeout = gloo_timeout
if listen_socket is not None:
store = create_tcp_store(