Skip to content

[Test] [AMD] diffusion: Skip the Kandinsky6 FA dispatch test on ROCm #44421

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

Merged
yichiche merged 4 commits into from
Oct 20, 2026
Merged

[Test] [AMD] diffusion: Skip the Kandinsky6 FA dispatch test on ROCm #44421

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
187ade9
[diffusion] fix: skip platform-specific attention asserts on ROCm in …
michaelzhang-ai Oct 15, 2026
b311e39
[diffusion] fix: pin is_rocm in the FastH3 server-args tests
michaelzhang-ai Oct 15, 2026
d354e81
[diffusion] fix: run the FastH3 ROCm-default test only on ROCm
michaelzhang-ai Oct 15, 2026
56a8c45
Merge main into fix-kandinsky6-sdpa-amd
michaelzhang-ai Oct 19, 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
7 changes: 6 additions & 1 deletion python/sglang/multimodal_gen/test/unit/test_kandinsky6_bf16_dtype.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,10 @@
from sglang.multimodal_gen.runtime.pipelines_core.stages.model_specific_stages.kandinsky6.denoising import (
Kandinsky6DenoisingStage,
)
from sglang.multimodal_gen.runtime.platforms import AttentionBackendEnum
from sglang.multimodal_gen.runtime.platforms import (
AttentionBackendEnum,
current_platform,
)
from sglang.multimodal_gen.runtime.server_args import get_global_server_args
from sglang.multimodal_gen.runtime.utils import precision

Expand Down Expand Up @@ -73,6 +76,8 @@ def test_bf16_projection_accepts_fp32_conditioning(modulation):
def test_attention_backend_dispatch_and_repeated_forward(monkeypatch, backend, lengths):
if backend == "FA" and torch.cuda.get_device_capability()[0] == 12:
pytest.skip("the platform currently resolves FA to SDPA on SM12.x")
if backend == "FA" and current_platform.is_rocm():
pytest.skip("the platform currently resolves FA to SDPA on ROCm")
if backend == "SAGE_ATTN_3":
if importlib.util.find_spec("sageattn3") is None:
pytest.skip("requires the optional SageAttention3 extension")
Expand Down
Loading