[video][bugfix] Accept semantic video output in offline examples (#6747)

Signed-off-by: david6666666 <530634352@qq.com>

examples/offline_inference/image_to_video/image_to_video.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
 
 """
 Image-to-Video generation example using Wan2.2 I2V/TI2V models, LTX2/LTX-2.3,
@@ -111,6 +111,13 @@ def build_image_to_video_prompt(
     return result
 
 
+def _validate_video_output_type(output_type: str) -> None:
+    if output_type not in {"image", "video"}:
+        raise ValueError(
+            f"Unexpected output type '{output_type}', expected 'video' or legacy 'image' for video generation."
+        )
+
+
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
         description=(
@@ -665,10 +672,7 @@ def main():
         frames = frames[0] if frames else None
 
     if isinstance(frames, OmniRequestOutput):
-        if frames.final_output_type != "image":
-            raise ValueError(
-                f"Unexpected output type '{frames.final_output_type}', expected 'image' for video generation."
-            )
+        _validate_video_output_type(frames.final_output_type)
         if frames.multimodal_output and "audio" in frames.multimodal_output:
             audio = frames.multimodal_output["audio"]
             audio_sample_rate = frames.multimodal_output.get("audio_sample_rate", audio_sample_rate)

examples/offline_inference/text_to_video/text_to_video.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
 
 import argparse
 import json
@@ -174,6 +174,13 @@ def build_text_to_video_prompt(prompt: str, negative_prompt: str | None) -> dict
     return result
 
 
+def _validate_video_output_type(output_type: str) -> None:
+    if output_type not in {"image", "video"}:
+        raise ValueError(
+            f"Unexpected output type '{output_type}', expected 'video' or legacy 'image' for video generation."
+        )
+
+
 def _normalize_float_tensor(tensor: torch.Tensor, source_range: str) -> torch.Tensor:
     """Normalize decoded frames according to the pipeline's output contract."""
     if not tensor.is_floating_point():
@@ -620,10 +627,7 @@ def main():
         frames = frames[0] if frames else None
 
     if isinstance(frames, OmniRequestOutput):
-        if frames.final_output_type != "image":
-            raise ValueError(
-                f"Unexpected output type '{frames.final_output_type}', expected 'image' for video generation."
-            )
+        _validate_video_output_type(frames.final_output_type)
         if frames.multimodal_output and "audio" in frames.multimodal_output:
             audio = frames.multimodal_output["audio"]
             audio_sample_rate = frames.multimodal_output.get("audio_sample_rate", audio_sample_rate)

tests/examples/offline_inference/test_image_task_prompts.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
 
 from __future__ import annotations
 
@@ -65,3 +65,38 @@ def test_image_to_video_builds_canonical_prompt() -> None:
         "multi_modal_data": {"image": image},
         "negative_prompt": "flicker",
     }
+
+
+@pytest.mark.parametrize(
+    ("relative_path", "module_name"),
+    [
+        ("offline_inference/text_to_video/text_to_video.py", "text_to_video_output_example"),
+        ("offline_inference/image_to_video/image_to_video.py", "image_to_video_output_example"),
+    ],
+)
+@pytest.mark.parametrize("output_type", ["video", "image"])
+def test_video_examples_accept_semantic_and_legacy_output_types(
+    relative_path: str,
+    module_name: str,
+    output_type: str,
+) -> None:
+    mod = _load_example_module(relative_path, module_name)
+
+    mod._validate_video_output_type(output_type)
+
+
+@pytest.mark.parametrize(
+    ("relative_path", "module_name"),
+    [
+        ("offline_inference/text_to_video/text_to_video.py", "text_to_video_invalid_output_example"),
+        ("offline_inference/image_to_video/image_to_video.py", "image_to_video_invalid_output_example"),
+    ],
+)
+def test_video_examples_reject_unrelated_output_types(
+    relative_path: str,
+    module_name: str,
+) -> None:
+    mod = _load_example_module(relative_path, module_name)
+
+    with pytest.raises(ValueError, match="Unexpected output type 'audio'"):
+        mod._validate_video_output_type("audio")