Skip to content

[BugFix][DP] Send DP wave completion only from dp_rank==0 #18502

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 22, 2025
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
[BugFix][DP] Send DP wave completion only from dp_rank==0
We were sending this notification back to the front-end for local_dp_rank==0 rather than dp_rank==0 which could result in redundant messages.

Since the wave count is included in the message explicitly, this doesn't actually cause any functional issues.

Co-authored-by: kourosh hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Nick Hill <nhill@redhat.com>
  • Loading branch information
njhill and kouroshHakha committed May 21, 2025
commit c1832a7b5221c7487b5e954be2c7eaf74632da6d
4 changes: 2 additions & 2 deletions vllm/v1/engine/core.py
Original file line number Diff line number Diff line change
Expand Up @@ -701,7 +701,7 @@ def _init_data_parallel(self, vllm_config: VllmConfig):
for i in range(local_dp_rank * world_size, (local_dp_rank + 1) *
world_size))

self.local_dp_rank = local_dp_rank
self.dp_rank = dp_rank
self.dp_group = vllm_config.parallel_config.stateless_init_dp_group()
self.current_wave = 0

Expand Down Expand Up @@ -774,7 +774,7 @@ def run_busy_loop(self):
local_unfinished_reqs)

if not self.engines_running:
if self.local_dp_rank == 0:
if self.dp_rank == 0:
# Notify client that we are pausing the loop.
logger.debug("Wave %d finished, pausing engine loop.",
self.current_wave)
Expand Down