Fix Stale Encoder Cache After Weight Update (#45093)

Signed-off-by: littlecircle0730 <littlecircle0730@gmail.com>

vllm/entrypoints/llm.py

@@ -898,6 +898,12 @@ class LLM(BeamSearchOfflineMixin, PoolingOfflineMixin, OfflineInferenceMixin):
     def finish_weight_update(self) -> None:
         """Finish the current weight update."""
         self.llm_engine.collective_rpc("finish_weight_update")
+        # Invalidate cached state computed with the old weights so it isn't
+        # reused for subsequent requests:
+        # - prefix cache: KV blocks computed with the old weights
+        # - encoder cache: multimodal embeddings keyed only by mm_hash
+        self.llm_engine.reset_prefix_cache()
+        self.llm_engine.reset_encoder_cache()
 
     def __repr__(self) -> str:
         """Return a transformers-style hierarchical view of the model."""

vllm/v1/engine/async_llm.py

@@ -1109,3 +1109,9 @@ class AsyncLLM(EngineClient):
     async def finish_weight_update(self) -> None:
         """Finish the current weight update."""
         await self.collective_rpc("finish_weight_update")
+        # Invalidate cached state computed with the old weights so it isn't
+        # reused for subsequent requests:
+        # - prefix cache: KV blocks computed with the old weights
+        # - encoder cache: multimodal embeddings keyed only by mm_hash
+        await self.reset_prefix_cache()
+        await self.reset_encoder_cache()