Skip to content
This repository has been archived by the owner on Oct 11, 2024. It is now read-only.

Commit

Permalink
format
Browse files Browse the repository at this point in the history
  • Loading branch information
robertgshaw2-neuralmagic committed Jun 9, 2024
1 parent 5335ad9 commit 611cfed
Show file tree
Hide file tree
Showing 2 changed files with 10 additions and 9 deletions.
3 changes: 2 additions & 1 deletion setup.py
Original file line number Diff line number Diff line change
Expand Up @@ -412,7 +412,8 @@ def _read_requirements(filename: str) -> List[str]:
_sparsity_deps = ["nm-magic-wand-nightly"]
nm_release_type = os.getenv(NM_RELEASE_TYPE)
if nm_release_type == 'RELEASE':
# gate magic-wand version in nm-vllm for release; for nightly, we always install the latest
# Gate magic-wand version in nm-vllm for release;
# For nightly, we always install the latest
magic_wand_version_dep = "0.2.2"
_sparsity_deps = [f"nm-magic-wand~={magic_wand_version_dep}"]

Expand Down
16 changes: 8 additions & 8 deletions tests/kernels/test_cutlass.py
Original file line number Diff line number Diff line change
Expand Up @@ -92,8 +92,8 @@ def cutlass_int8_gemm_helper(m: int,
# automation system yet.
@pytest.mark.skipif(capability < 90,
reason="FP8 cutlass is not supported on this GPU "
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
def test_cutlass_fp8_gemm(m: int, n: int, k: int, per_act_token: bool,
per_out_ch: bool):
cutlass_fp8_gemm_helper(m, n, k, per_act_token, per_out_ch)
Expand Down Expand Up @@ -126,8 +126,8 @@ def test_cutlass_int8_gemm_output_dtype(per_act_token: bool, per_out_ch: bool,
# automation system yet.
@pytest.mark.skipif(capability < 90,
reason="FP8 cutlass is not supported on this GPU "
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
def test_cutlass_fp8_gemm_output_dtype(per_act_token: bool, per_out_ch: bool,
out_dtype: Type[torch.dtype]):
cutlass_fp8_gemm_helper(512, 512, 512, per_act_token, per_out_ch,
Expand All @@ -142,8 +142,8 @@ def test_cutlass_fp8_gemm_output_dtype(per_act_token: bool, per_out_ch: bool,
# automation system yet.
@pytest.mark.skipif(capability < 90,
reason="FP8 cutlass is not supported on this GPU "
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
def test_cutlass_fp8_gemm_devices(per_act_token: bool, per_out_ch: bool,
device: str):
cutlass_fp8_gemm_helper(512, 512, 512, per_act_token, per_out_ch,
Expand Down Expand Up @@ -171,8 +171,8 @@ def test_cutlass_int8_gemm_devices(per_act_token: bool, per_out_ch: bool,
# automation system yet.
@pytest.mark.skipif(capability < 90,
reason="FP8 cutlass is not supported on this GPU "
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
"type because we need CUDA 12.4 + we do "
"not have this in automation yet.")
def test_cutlass_fp8_gemm_m_sweep(per_act_token: bool, per_out_ch: bool):
for nk in range(32, 128, 32):
for m in range(1, 128):
Expand Down

0 comments on commit 611cfed

Please sign in to comment.