[Bugfix] reasoning_parser parameter handling in run_batch.py (#26225)

Signed-off-by: inc-jeong <inc.jeong@navercorp.com>
Signed-off-by: InChang Jeong <inc.jeong@navercorp.com>
Co-authored-by: USER <user@AL02367916.local>

tests/entrypoints/openai/test_run_batch.py

@@ -39,6 +39,9 @@ INPUT_RERANK_BATCH = """{"custom_id": "request-1", "method": "POST", "url": "/re
 {"custom_id": "request-2", "method": "POST", "url": "/v1/rerank", "body": {"model": "BAAI/bge-reranker-v2-m3", "query": "What is the capital of France?", "documents": ["The capital of Brazil is Brasilia.", "The capital of France is Paris."]}}
 {"custom_id": "request-2", "method": "POST", "url": "/v2/rerank", "body": {"model": "BAAI/bge-reranker-v2-m3", "query": "What is the capital of France?", "documents": ["The capital of Brazil is Brasilia.", "The capital of France is Paris."]}}"""
 
+INPUT_REASONING_BATCH = """{"custom_id": "request-1", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "Qwen/Qwen3-0.6B", "messages": [{"role": "system", "content": "You are a helpful assistant."},{"role": "user", "content": "Solve this math problem: 2+2=?"}]}}
+{"custom_id": "request-2", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "Qwen/Qwen3-0.6B", "messages": [{"role": "system", "content": "You are a helpful assistant."},{"role": "user", "content": "What is the capital of France?"}]}}"""
+
 
 def test_empty_file():
     with (
@@ -188,3 +191,50 @@ def test_score(input_batch):
             line_dict = json.loads(line)
             assert isinstance(line_dict, dict)
             assert line_dict["error"] is None
+
+
+def test_reasoning_parser():
+    """
+    Test that reasoning_parser parameter works correctly in run_batch.
+    """
+    with (
+        tempfile.NamedTemporaryFile("w") as input_file,
+        tempfile.NamedTemporaryFile("r") as output_file,
+    ):
+        input_file.write(INPUT_REASONING_BATCH)
+        input_file.flush()
+        proc = subprocess.Popen(
+            [
+                "vllm",
+                "run-batch",
+                "-i",
+                input_file.name,
+                "-o",
+                output_file.name,
+                "--model",
+                "Qwen/Qwen3-0.6B",
+                "--reasoning-parser",
+                "qwen3",
+            ],
+        )
+        proc.communicate()
+        proc.wait()
+        assert proc.returncode == 0, f"{proc=}"
+
+        contents = output_file.read()
+        for line in contents.strip().split("\n"):
+            # Ensure that the output format conforms to the openai api.
+            # Validation should throw if the schema is wrong.
+            BatchRequestOutput.model_validate_json(line)
+
+            # Ensure that there is no error in the response.
+            line_dict = json.loads(line)
+            assert isinstance(line_dict, dict)
+            assert line_dict["error"] is None
+
+            # Check that reasoning_content is present and not empty
+            reasoning_content = line_dict["response"]["body"]["choices"][0]["message"][
+                "reasoning_content"
+            ]
+            assert reasoning_content is not None
+            assert len(reasoning_content) > 0

vllm/entrypoints/openai/run_batch.py

@@ -31,6 +31,7 @@ from vllm.entrypoints.openai.serving_embedding import OpenAIServingEmbedding
 from vllm.entrypoints.openai.serving_models import BaseModelPath, OpenAIServingModels
 from vllm.entrypoints.openai.serving_score import ServingScores
 from vllm.logger import init_logger
+from vllm.reasoning import ReasoningParserManager
 from vllm.utils import FlexibleArgumentParser, random_uuid
 from vllm.version import __version__ as VLLM_VERSION
 
@@ -331,6 +332,17 @@ async def run_request(
     return batch_output
 
 
+def validate_run_batch_args(args):
+    valid_reasoning_parses = ReasoningParserManager.reasoning_parsers.keys()
+    if (
+        reasoning_parser := args.structured_outputs_config.reasoning_parser
+    ) and reasoning_parser not in valid_reasoning_parses:
+        raise KeyError(
+            f"invalid reasoning parser: {reasoning_parser} "
+            f"(chose from {{ {','.join(valid_reasoning_parses)} }})"
+        )
+
+
 async def run_batch(
     engine_client: EngineClient,
     args: Namespace,
@@ -359,6 +371,7 @@ async def run_batch(
         base_model_paths=base_model_paths,
         lora_modules=None,
     )
+
     openai_serving_chat = (
         OpenAIServingChat(
             engine_client,
@@ -367,12 +380,14 @@ async def run_batch(
             request_logger=request_logger,
             chat_template=None,
             chat_template_content_format="auto",
+            reasoning_parser=args.structured_outputs_config.reasoning_parser,
             enable_prompt_tokens_details=args.enable_prompt_tokens_details,
             enable_force_include_usage=args.enable_force_include_usage,
         )
         if "generate" in supported_tasks
         else None
     )
+
     openai_serving_embedding = (
         OpenAIServingEmbedding(
             engine_client,
@@ -504,6 +519,8 @@ async def main(args: Namespace):
     from vllm.entrypoints.openai.api_server import build_async_engine_client
     from vllm.usage.usage_lib import UsageContext
 
+    validate_run_batch_args(args)
+
     async with build_async_engine_client(
         args,
         usage_context=UsageContext.OPENAI_BATCH_RUNNER,