From d5fd9eb6bbeb067582a42f7bd060996c7c6931c0 Mon Sep 17 00:00:00 2001 From: wangyu <53896905+yenuo26@users.noreply.github.com> Date: Wed, 2 Sep 2026 23:16:58 +0800 Subject: [PATCH] [Bugfix][Diffusion] Register Sana I2V pipeline as video output (#6953) Signed-off-by: wangyu <410167048@qq.com> --- .../openai_api/test_video_pipeline_capability.py | 11 +++++++++++ vllm_omni/diffusion/model_metadata.py | 1 + 2 files changed, 12 insertions(+) diff --git a/tests/entrypoints/openai_api/test_video_pipeline_capability.py b/tests/entrypoints/openai_api/test_video_pipeline_capability.py index cfa11f96d..985aadf89 100644 --- a/tests/entrypoints/openai_api/test_video_pipeline_capability.py +++ b/tests/entrypoints/openai_api/test_video_pipeline_capability.py @@ -64,3 +64,14 @@ def test_video_pipeline_rejects_non_video_final_outputs(stage_configs): ) def test_registered_video_aliases_declare_video_output(model_class_name): assert get_diffusion_output_type(model_class_name) == "video" + + +@pytest.mark.parametrize( + "model_class_name", + [ + "SanaVideoPipeline", + "SanaImageToVideoPipeline", + ], +) +def test_sana_video_pipelines_declare_video_output(model_class_name): + assert get_diffusion_output_type(model_class_name) == "video" diff --git a/vllm_omni/diffusion/model_metadata.py b/vllm_omni/diffusion/model_metadata.py index 1665d6a16..7f15955ec 100644 --- a/vllm_omni/diffusion/model_metadata.py +++ b/vllm_omni/diffusion/model_metadata.py @@ -88,6 +88,7 @@ _DIFFUSION_MODEL_METADATA: dict[str, DiffusionModelMetadata] = { "Cosmos3OmniDiffusersPipeline": DiffusionModelMetadata(final_output_type="video"), "Cosmos3OmniPipeline": DiffusionModelMetadata(final_output_type="video"), "SanaVideoPipeline": DiffusionModelMetadata(final_output_type="video"), + "SanaImageToVideoPipeline": DiffusionModelMetadata(final_output_type="video"), "SanaWmPipeline": DiffusionModelMetadata( supports_multimodal_inputs=True, max_multimodal_image_inputs=1,