From f60c6b33a5c637315624c841775b5efdf6fb46d6 Mon Sep 17 00:00:00 2001 From: Vadim Gimpelson <156319763+vadiklyutiy@users.noreply.github.com> Date: Thu, 14 May 2026 19:39:24 +0400 Subject: [PATCH] [V1][DP][LB] Publish request counts at the start of each engine step (#41626) Signed-off-by: Vadim Gimpelson Signed-off-by: Vadim Gimpelson <156319763+vadiklyutiy@users.noreply.github.com> Co-authored-by: Nick Hill --- vllm/v1/engine/core.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/vllm/v1/engine/core.py b/vllm/v1/engine/core.py index d19d928c70..439848ea42 100644 --- a/vllm/v1/engine/core.py +++ b/vllm/v1/engine/core.py @@ -1817,6 +1817,8 @@ class DPEngineCoreProc(EngineCoreProc): while self._handle_shutdown(): # 1) Poll the input queue until there is work to do. self._process_input_queue() + # Publish request counts before and after GPU step to ensure freshness. + self._maybe_publish_request_counts() if self.eep_scaling_state is not None: _ = self.eep_scaling_state.progress()