diff --git a/python/sglang/srt/disaggregation/mini_lb.py b/python/sglang/srt/disaggregation/mini_lb.py index fdc9f2f45..55db8fa03 100644 --- a/python/sglang/srt/disaggregation/mini_lb.py +++ b/python/sglang/srt/disaggregation/mini_lb.py @@ -232,6 +232,9 @@ async def handle_generate_request(request_data: dict): yield b"data: " + orjson.dumps( error_msg, option=orjson.OPT_NON_STR_KEYS ) + b"\n\n" + finally: + if prefill_response is not None: + await prefill_response.release() return StreamingResponse( stream_results(),