[multi-step] Raise error if not using async engine (vllm-project#7703)

abmfy · Aug 21, 2024 · 91f4522 · 91f4522
1 parent 1b32e02
commit 91f4522
Showing 1 changed file with 5 additions and 0 deletions.
diff --git a/vllm/engine/llm_engine.py b/vllm/engine/llm_engine.py
@@ -1302,6 +1302,11 @@ def step(self) -> List[Union[RequestOutput, EmbeddingRequestOutput]]:
             raise NotImplementedError(
                 "Pipeline parallelism is only supported through AsyncLLMEngine "
                 "as performance will be severely degraded otherwise.")
+
+        if self.scheduler_config.num_scheduler_steps > 1:
+            raise NotImplementedError(
+                "Multiple scheduler steps (multi-step) are only supported "
+                "through AsyncLLMEngine. ")
         seq_group_metadata_list, scheduler_outputs = self.scheduler[
             0].schedule()