diff --git a/python/sglang/srt/managers/tokenizer_manager.py b/python/sglang/srt/managers/tokenizer_manager.py
index 25c1bb3f7..41f92e299 100644
--- a/python/sglang/srt/managers/tokenizer_manager.py
+++ b/python/sglang/srt/managers/tokenizer_manager.py
@@ -996,7 +996,11 @@ class TokenizerManager(TokenizerCommunicatorMixin):
                         finish_reason.get("type") == "abort"
                         and finish_reason.get("status_code") == HTTPStatus.BAD_REQUEST
                     ):
-                        raise ValueError(finish_reason["message"])
+                        if not obj.stream:
+                            raise ValueError(finish_reason["message"])
+                        else:
+                            yield out
+                            break
 
                     if finish_reason.get("type") == "abort" and finish_reason.get(
                         "status_code"
@@ -1013,11 +1017,14 @@ class TokenizerManager(TokenizerCommunicatorMixin):
                         # Mark ongoing LoRA request as finished.
                         if self.server_args.enable_lora and state.obj.lora_path:
                             await self.lora_registry.release(state.obj.lora_id)
-
-                        raise fastapi.HTTPException(
-                            status_code=finish_reason["status_code"],
-                            detail=finish_reason["message"],
-                        )
+                        if not obj.stream:
+                            raise fastapi.HTTPException(
+                                status_code=finish_reason["status_code"],
+                                detail=finish_reason["message"],
+                            )
+                        else:
+                            yield out
+                            break
                 yield out
                 break