[Bugfix] Stop reading the removed OmniRequestOutput.request_output accessor (#6172)

Signed-off-by: MrlixiangWE <mrdanaer@gmail.com>

docs/features/custom_pipeline.md

@@ -101,7 +101,7 @@ outputs = omni.generate(
 )
 
 # Access custom trajectory data
-output = outputs[0].request_output
+output = outputs[0]
 print(f"Trajectory timesteps shape: {output.metrics['trajectory_timesteps'].shape}")
 print(f"Trajectory latents shape: {output.latents.shape}")
 ```

docs/getting_started/quickstart.md

@@ -47,7 +47,7 @@ if __name__ == "__main__":
     omni = Omni(model="Tongyi-MAI/Z-Image-Turbo")
     prompt = "a cup of coffee on the table"
     outputs = omni.generate(prompt)
-    images = outputs[0].request_output.images
+    images = outputs[0].images
     images[0].save("coffee.png")
 ```
 
@@ -75,8 +75,7 @@ if __name__ == "__main__":
     ]
     omni_outputs = omni.generate(prompts)
     for i_prompt, prompt_output in enumerate(omni_outputs):
-        this_request_output = prompt_output.request_output
-        this_images = this_request_output.images
+        this_images = prompt_output.images
         for i_image, image in enumerate(this_images):
             image.save(f"p{i_prompt}-img{i_image}.jpg")
             print("saved to", f"p{i_prompt}-img{i_image}.jpg")

docs/user_guide/examples/offline_inference/glm_image.md

@@ -43,7 +43,7 @@ if __name__ == "__main__":
             ),
         ],
     )
-    outputs[0].request_output.images[0].save("output.png")
+    outputs[0].images[0].save("output.png")
 ```
 
 ## Image-to-Image (Image Editing)
@@ -72,7 +72,7 @@ if __name__ == "__main__":
             ),
         ],
     )
-    outputs[0].request_output.images[0].save("output.png")
+    outputs[0].images[0].save("output.png")
 ```
 
 ## Generation Parameters

docs/user_guide/examples/offline_inference/text_to_image.md

@@ -55,7 +55,7 @@ if __name__ == "__main__":
     omni = Omni(model="Qwen/Qwen-Image")
     prompt = "a cup of coffee on the table"
     outputs = omni.generate(prompt)
-    images = outputs[0].request_output.images
+    images = outputs[0].images
     images[0].save("coffee.png")
 ```
 
@@ -179,7 +179,7 @@ if __name__ == "__main__":
     ]
     outputs = omni.generate(prompts)
     for i, output in enumerate(outputs):
-        output.request_output.images[0].save(f"{i}.jpg")
+        output.images[0].save(f"{i}.jpg")
 ```
 
 !!! info
@@ -207,7 +207,7 @@ if __name__ == "__main__":
         }
     ])
     for i, output in enumerate(outputs):
-        output.request_output.images[0].save(f"{i}.jpg")
+        output.images[0].save(f"{i}.jpg")
 ```
 
 You can also pass a negative prompt via the CLI argument `--negative-prompt`:

examples/offline_inference/image_to_video/README.md

@@ -96,7 +96,7 @@ if __name__ == "__main__":
         ),
     )
     from diffusers.utils import export_to_video
-    frames = outputs[0].request_output.images
+    frames = outputs[0].images
     export_to_video(frames, "quick_test_i2v.mp4", fps=16)
 ```
 

examples/offline_inference/minicpmo/end2end.py

@@ -335,7 +335,7 @@ def main(args):
     print(f"use_tts (tts_bos): {use_tts}")
 
     for stage_outputs in omni_generator:
-        output = stage_outputs.request_output
+        output = stage_outputs
         if stage_outputs.final_output_type == "text":
             request_id = output.request_id
             text_output = output.outputs[0].text

examples/offline_inference/text_to_image/README.md

@@ -74,7 +74,7 @@ if __name__ == "__main__":
     omni = Omni(model="Qwen/Qwen-Image")
     prompt = "a cup of coffee on the table"
     outputs = omni.generate(prompt)
-    images = outputs[0].request_output.images
+    images = outputs[0].images
     images[0].save("coffee.png")
 ```
 
@@ -265,7 +265,7 @@ if __name__ == "__main__":
     ]
     outputs = omni.generate(prompts)
     for i, output in enumerate(outputs):
-        output.request_output.images[0].save(f"{i}.jpg")
+        output.images[0].save(f"{i}.jpg")
 ```
 
 !!! info
@@ -301,7 +301,7 @@ if __name__ == "__main__":
         }
     ])
     for i, output in enumerate(outputs):
-        output.request_output.images[0].save(f"{i}.jpg")
+        output.images[0].save(f"{i}.jpg")
 ```
 
 You can also pass a negative prompt via the CLI argument `--negative-prompt`:

tests/diffusion/distributed/test_flux2_vae_patch_parallel.py

@@ -53,14 +53,10 @@ def _diff_metrics(a: Image.Image, b: Image.Image) -> tuple[float, float]:
 def _extract_single_image(outputs) -> Image.Image:
     first_output = outputs[0]
     assert first_output.final_output_type == "image"
-    if not hasattr(first_output, "request_output") or not first_output.request_output:
-        raise ValueError("No request_output found in OmniRequestOutput")
+    if not isinstance(first_output, OmniRequestOutput):
+        raise ValueError(f"Expected an OmniRequestOutput, got {type(first_output).__name__}")
 
-    req_out = first_output.request_output
-    if not isinstance(req_out, OmniRequestOutput) or not hasattr(req_out, "images"):
-        raise ValueError("Invalid request_output structure or missing 'images' key")
-
-    images = req_out.images
+    images = first_output.images
     if images is None or len(images) != 1:
         raise ValueError(f"Expected 1 image, got {0 if images is None else len(images)}")
     return images[0]

tests/diffusion/distributed/test_vae_decode_parallelism.py

@@ -72,8 +72,7 @@ def _to_float_array(data: Any) -> np.ndarray:
 
 
 def _extract_output_array(output: Any) -> np.ndarray:
-    req_out = output.request_output
-    payload = req_out.images[0]
+    payload = output.images[0]
     return _to_float_array(payload)
 
 

tests/diffusion/models/minimax_h3/test_minimax_h3_quantization_quality.py

@@ -142,12 +142,6 @@ def _generate_joint_output(omni, config: QualityTestConfig):
 
     first = outputs[0]
     worker_peak_memory_mb = float(getattr(first, "peak_memory_mb", 0.0) or 0.0)
-    if hasattr(first, "request_output") and isinstance(first.request_output, list):
-        first = first.request_output[0]
-        worker_peak_memory_mb = max(
-            worker_peak_memory_mb,
-            float(getattr(first, "peak_memory_mb", 0.0) or 0.0),
-        )
     if not hasattr(first, "images") or not first.images:
         raise ValueError("Could not extract video frames from H3 output")
     frames = first.images[0]

tests/docs/test_example_output_accessors.py

@@ -0,0 +1,55 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Guard the public accessor that examples and docs use to read omni outputs.
+
+``OmniRequestOutput`` used to hold a nested ``RequestOutput`` in a
+``request_output`` field.  Since #5146 it *inherits* ``RequestOutput``, so the
+generation content lives on the object itself and the nested field is gone.
+
+Call sites that kept the old accessor only fail once a real model runs, which
+means a GPU nightly job is the first thing to notice.  These checks pin both
+halves of the contract on CPU instead: the field is really gone, and no doc,
+example, test or tool still reaches for it.
+"""
+
+import re
+from pathlib import Path
+
+import pytest
+
+from vllm_omni.outputs import OmniRequestOutput
+
+pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
+
+
+ROOT_DIR = Path(__file__).parents[2]
+SCANNED_DIRS = ("docs", "examples", "tests", "tools")
+SCANNED_SUFFIXES = (".py", ".md")
+
+# ``.request_outputs`` (plural, an OutputProcessor batch) is a different and
+# still-valid attribute, so require a non-identifier character after the name.
+NESTED_ACCESSOR = re.compile(r"\.request_output(?![\w])")
+
+
+def _iter_scanned_files():
+    for directory in SCANNED_DIRS:
+        for path in sorted((ROOT_DIR / directory).rglob("*")):
+            if path.suffix in SCANNED_SUFFIXES and path.resolve() != Path(__file__).resolve():
+                yield path
+
+
+def test_omni_request_output_has_no_nested_request_output():
+    assert not hasattr(OmniRequestOutput(request_id="req-accessor"), "request_output")
+
+
+def test_no_source_reads_the_nested_request_output():
+    offenders = []
+    for path in _iter_scanned_files():
+        for lineno, line in enumerate(path.read_text(encoding="utf-8").splitlines(), start=1):
+            if NESTED_ACCESSOR.search(line):
+                offenders.append(f"{path.relative_to(ROOT_DIR)}:{lineno}: {line.strip()}")
+
+    assert not offenders, (
+        "OmniRequestOutput no longer nests a RequestOutput; read the content off the "
+        "output itself (e.g. `outputs[0].images`) instead of `.request_output`:\n" + "\n".join(offenders)
+    )

tools/baselines/omni_hash_baseline.py

@@ -159,11 +159,10 @@ def _run_once(model: str, yaml_path: str, *, async_mode: bool) -> list[dict]:
             wav: bytes = b""
             # Batch size 1: submit a single prompt so scheduling is deterministic.
             for step in omni.generate([_text_prompt(question)], _sampling_params(), py_generator=False):
-                out = step.request_output
                 if step.final_output_type == "text":
-                    text = out.outputs[0].text  # exact text, no strip -> truly bit-identical
+                    text = step.outputs[0].text  # exact text, no strip -> truly bit-identical
                 elif step.final_output_type == "audio":
-                    wav = _audio_bytes(out.outputs[0].multimodal_output["audio"])
+                    wav = _audio_bytes(step.outputs[0].multimodal_output["audio"])
             if text is None:
                 raise RuntimeError(f"prompt_id={prompt_id} async={async_mode}: produced no text output")
             if not wav: