[Bugfix] Mark image edit benchmark stream errors as failed (#8015)

Signed-off-by: levius <2114377220@qq.com>

tests/benchmarks/patch/test_patch.py

@@ -1476,6 +1476,42 @@ async def test_image_edits_stream_true_uses_sse_path(mocker: MockerFixture) -> N
     assert output.image_count == 1
 
 
+@pytest.mark.asyncio
+async def test_image_edits_stream_error_marks_request_failed(mocker: MockerFixture) -> None:
+    """HTTP 200 image-edit streams can terminate with an error event."""
+    sse_chunk = (
+        b'data: {"object":"image.edit.chunk","type":"ar_delta","delta":"partial"}\n\n'
+        b'data: {"object":"error","error":{"message":"image generation failed",'
+        b'"type":"server_error","code":500}}\n\n'
+        b"data: [DONE]\n\n"
+    )
+    mock_response = MockResponse(200, [sse_chunk])
+    mock_session = mocker.AsyncMock()
+    mock_session.post = mocker.MagicMock(return_value=mock_response)
+
+    request = RequestFuncInput(
+        model="multi-stage-edit",
+        model_name="multi-stage-edit",
+        prompt="edit",
+        api_url="http://test.com/v1/images/edits",
+        prompt_len=2,
+        output_len=1,
+        multi_modal_content=[
+            {
+                "type": "image_url",
+                "image_url": {"url": f"data:image/png;base64,{_MIN_PNG_B64}"},
+            }
+        ],
+        extra_body={"stream": True},
+    )
+
+    output = await async_request_openai_image_edits_omni(request, mock_session, pbar=None)
+
+    assert output.success is False
+    assert output.error == "image generation failed"
+    assert output.generated_text == "partial"
+
+
 @pytest.mark.asyncio
 async def test_image_generations_e2el_includes_json_body_consume(mocker: MockerFixture) -> None:
     """E2EL must include body transfer/decode, not stop at HTTP headers."""

vllm_omni/benchmarks/patch/patch.py

@@ -2409,6 +2409,7 @@ async def async_request_openai_image_edits_omni(
                 timestamp = st
                 most_recent_text_timestamp = st
                 generated_text = ""
+                streaming_error_received = False
                 handler = StreamedResponseHandler()
                 async for chunk_bytes in response.content.iter_any():
                     if not chunk_bytes:
@@ -2424,6 +2425,13 @@ async def async_request_openai_image_edits_omni(
 
                         timestamp = time.perf_counter()
                         data = json.loads(chunk)
+                        if (streaming_error := data.get("error")) is not None:
+                            streaming_error_received = True
+                            if isinstance(streaming_error, dict):
+                                output.error = str(streaming_error.get("message") or streaming_error)
+                            else:
+                                output.error = str(streaming_error)
+                            continue
                         _update_output_stage_metrics_from_payload(
                             output,
                             data,
@@ -2464,7 +2472,7 @@ async def async_request_openai_image_edits_omni(
                             output.denoise_step_latency_ms = metrics_denoise_step_ms
                 output.latency = timestamp - st
                 output.generated_text = generated_text
-                output.success = True
+                output.success = not streaming_error_received
             else:
                 data = await response.json()
                 _finalize_image_json_http_response(