[Bugfix] Return HTTP 500 for non-streaming generate errors (#49195)

Signed-off-by: WEI CHENG CHIU <waynehacking8@gmail.com>

tests/entrypoints/scale_out/token_in_token_out/test_generate_stream.py

@@ -9,7 +9,7 @@ from unittest.mock import AsyncMock, MagicMock
 import pytest
 
 from vllm.config.multimodal import MultiModalConfig
-from vllm.entrypoints.openai.engine.protocol import StreamOptions
+from vllm.entrypoints.openai.engine.protocol import GenerationError, StreamOptions
 from vllm.entrypoints.openai.models.protocol import BaseModelPath
 from vllm.entrypoints.openai.models.serving import OpenAIServingModels
 from vllm.entrypoints.scale_out.token_in_token_out.protocol import (
@@ -229,6 +229,30 @@ async def test_serve_tokens_threads_session_id_header_to_engine():
     assert engine.generate.call_args.kwargs["session_id"] == "header-session"
 
 
+@pytest.mark.asyncio
+async def test_non_stream_error():
+    """finish_reason='error' raises for the HTTP exception handler."""
+    engine = _mock_engine()
+
+    async def mock_generate(*args, **kwargs):
+        yield _make_request_output(
+            "req-1", token_ids=[], finish_reason="error", finished=True
+        )
+
+    engine.generate = MagicMock(side_effect=mock_generate)
+    serving = _build_serving_tokens(engine)
+
+    request = GenerateRequest(
+        token_ids=[1, 2, 3],
+        sampling_params=SamplingParams(max_tokens=10),
+        model=MODEL_NAME,
+        stream=False,
+    )
+
+    with pytest.raises(GenerationError):
+        await serving.serve_tokens(request)
+
+
 @pytest.mark.asyncio
 async def test_stream_basic():
     """Streaming returns SSE chunks with correct token_ids and ends with [DONE]."""

vllm/entrypoints/scale_out/token_in_token_out/serving.py

@@ -288,6 +288,8 @@ class ServingTokens(GenerateBaseServing):
         choices: list[GenerateResponseChoice] = []
         num_generated_tokens = 0
         for output in final_res.outputs:
+            self._raise_if_error(output.finish_reason, request_id)
+
             token_ids = output.token_ids
             out_logprobs = output.logprobs