feat: add flush cache to EngineBase and HttpServerEngineAdapter (#6009)

python/sglang/srt/entrypoints/EngineBase.py

@@ -27,6 +27,11 @@ class EngineBase(ABC):
         """Generate outputs based on given inputs."""
         pass
 
+    @abstractmethod
+    def flush_cache(self):
+        """Flush the cache of the engine."""
+        pass
+
     @abstractmethod
     def update_weights_from_tensor(
         self,

python/sglang/srt/entrypoints/http_server_engine.py

@@ -140,3 +140,6 @@ class HttpServerEngineAdapter(EngineBase):
 
     def resume_memory_occupation(self):
         return self._make_request("resume_memory_occupation")
+
+    def flush_cache(self):
+        return self._make_request("flush_cache")