Skip to content

skip nvfp4 gemm #5758

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

Closed
liqiangxl wants to merge 3 commits into from
Closed

skip nvfp4 gemm #5758

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4b7d91b
fix test_grouped_mm
liqiangxl Jan 5, 2026
235f161
skip nvfp4
liqiangxl Jan 5, 2026
33d88bf
Merge branch 'main' into llu/skip_nvfp4_gemm
liqiangxl Jan 5, 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
21 changes: 13 additions & 8 deletions tests/python/direct/test_cutlass_nvfp4_gemm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,15 +6,8 @@
import pytest
import torch
from nvfuser_direct import nvf_cutlass

compute_cap = torch.cuda.get_device_capability()
if compute_cap < (10, 0) or compute_cap >= (12, 0):
pytest.skip(
reason="Nvfp4 Requires compute capability 10.",
allow_module_level=True,
)

from python.direct_utils import (
microarchitecture_is,
FLOAT4_E2M1_MAX,
FLOAT8_E4M3_MAX,
dequantize_to_dtype,
Expand Down Expand Up @@ -52,6 +45,10 @@ def get_ref_results(
return torch.matmul(a_in_dtype, b_in_dtype.t())


@pytest.mark.skipif(
not microarchitecture_is(10, 0),
reason="Does not support blackwell compute 12.0, other arches are not tested.",
)
@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16])
@pytest.mark.parametrize(
"shape", [(128, 128, 64), (128, 128, 128), (256, 128, 64), (128, 256, 128)]
Expand Down Expand Up @@ -100,6 +97,10 @@ def test_nvfp4_gemm(
torch.testing.assert_close(out, expected_out.to(dtype=dtype), atol=1e-1, rtol=1e-1)


@pytest.mark.skipif(
not microarchitecture_is(10, 0),
reason="Does not support blackwell compute 12.0, other arches are not tested.",
)
@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16])
@pytest.mark.parametrize(
"shape", [(128, 128, 64), (128, 128, 128), (256, 128, 64), (128, 256, 128)]
Expand Down Expand Up @@ -175,6 +176,10 @@ def test_nvfp4_gemm_epilogue(
)


@pytest.mark.skipif(
not microarchitecture_is(10, 0),
reason="Does not support blackwell compute 12.0, other arches are not tested.",
)
@pytest.mark.parametrize("config", [[1024, 128, 256]])
@pytest.mark.parametrize("tokens_per_expert_neg_one", [[115, 144, 8]])
@pytest.mark.parametrize("out_dtype", [torch.bfloat16, torch.float16])
Expand Down