Skip to content

[ci] skip bf16 precision test on T4. #15091

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sayakpaul wants to merge 2 commits into
base: main
Choose a base branch
from
+10 −0
Open

[ci] skip bf16 precision test on T4. #15091

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
be5c106
skip bf16 precision test on T4.
sayakpaul Oct 21, 2026
8b1bb98
Merge branch 'main' into gate-test_half_precision_inference_no_nan
sayakpaul Oct 21, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions tests/pipelines/stable_diffusion_xl/test_stable_diffusion_xl_img2img.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@
backend_empty_cache,
floats_tensor,
load_image,
require_accelerator,
slow,
torch_device,
)
Expand Down Expand Up @@ -184,6 +185,15 @@ def get_dummy_inputs(self):


class TestStableDiffusionXLImg2ImgPipeline(StableDiffusionXLImg2ImgPipelineTesterConfig, PipelineTesterMixin):
# Re-declares the base method's decorators, which overriding would otherwise drop.
@pytest.mark.skipif(torch_device not in ["cuda", "xpu"], reason="half-precision inference requires CUDA or XPU")
@require_accelerator
@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16], ids=str)
def test_half_precision_inference_no_nan(self, dtype):
if dtype == torch.bfloat16 and torch_device == "cuda" and torch.cuda.get_device_capability() < (8, 0):
pytest.xfail("produces NaNs in bfloat16 on GPUs without native bfloat16 support (e.g. T4)")
super().test_half_precision_inference_no_nan(dtype)

def test_components_function(self):
# `requires_aesthetics_score` is a config value, not a component, so it is not part of `pipe.components`.
init_components = self.get_dummy_components()
Expand Down
Loading