[BugFix] Fix multi-node offline data parallel (#19937)

Signed-off-by: Nick Hill <nhill@redhat.com>
2025-06-24 12:45:20 -07:00
parent c635c5f744
commit 8619e7158c
5 changed files with 31 additions and 4 deletions
--- a/vllm/v1/engine/llm_engine.py
+++ b/vllm/v1/engine/llm_engine.py
@@ -160,7 +160,7 @@ class LLMEngine:
    def has_unfinished_requests(self) -> bool:
        has_unfinished = self.output_processor.has_unfinished_requests()
        if self.dp_group is None:
-            return has_unfinished
+            return has_unfinished or self.engine_core.dp_engines_running()
        return self.has_unfinished_requests_dp(has_unfinished)

    def has_unfinished_requests_dp(self, has_unfinished: bool) -> bool: