[tests] remove test_pipelines_common.py - #14669
Conversation
| # support `"pt"`; adding that here is a `src/` change and out of scope for this test migration, so the affected | ||
| # tests are marked `xfail` rather than skipped: whoever adds the `"pt"` branch will see them XPASS and can drop | ||
| # these markers. | ||
| NO_PT_OUTPUT = pytest.mark.xfail( |
There was a problem hiding this comment.
Open to other ideas but I chose to xfail these tests and not introduce anything really new to our shared tester mixins.
|
/diffusers-bot pytest tests/pipelines/stable_diffusion_adapter/test_stable_diffusion_adapter.py |
|
✅ |
| return self.get_pipeline() | ||
|
|
||
| def enable_group_offload_on_component(pipe, group_offloading_kwargs): | ||
| offloaded_component_names = set() |
There was a problem hiding this comment.
| # `transformer`/`unet`/`controlnet`/`adapter`, and only puts `vae`/`vqvae`/`image_encoder` back on the accelerator. | ||
| # A prior pipeline's denoiser is called `prior`, so it matches neither list and is left on CPU while the text | ||
| # encoder is onloaded. Fixing this means widening the mixin's component lists, not changing the pipeline. | ||
| COMPONENT_GROUP_OFFLOAD_XFAIL_REASON = ( |
There was a problem hiding this comment.
Side effect of fixes in tests/pipelines/testing_utils/memory.py.
|
/diffusers-bot pytest tests/pipelines/stable_diffusion_2/test_stable_diffusion_depth.py::TestStableDiffusionDepth2ImgPipelineMemory::test_group_offloading_inference |
|
|
|
/diffusers-bot pytest tests/pipelines/consisid/test_consisid.py tests/pipelines/stable_audio/test_stable_audio.py |
|
✅ |
|
/diffusers-bot pytest tests/pipelines/stable_diffusion_2/test_stable_diffusion_depth.py::TestStableDiffusionDepth2ImgPipelineMemory::test_group_offloading_inference |
|
✅ |
1 participant
FluxIPAdapterTesterMixinto the IP Adapter testing utils since it's shared by more than Flux.test_pipelines_common.py.