biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

kernels/moe test pruning (#27053)

Browse Source 
Signed-off-by: Fardin Hoque <kfhfar@amazon.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
This commit is contained in:
Fardin Hoque
2025-10-29 21:10:34 -07:00
committed by GitHub
parent 17d055f527
commit b8c48c5d72
13 changed files with 34 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/kernels/moe/test_flashinfer_moe.py
View File

@@ -34,10 +34,8 @@ if not has_flashinfer_cutlass_fused_moe() or not current_platform.has_device_cap
MNK_FACTORS = [
(2, 1024, 1024),
(2, 1024, 1536),
(2, 3072, 1024),
(2, 3072, 1536),
(64, 1024, 1024),
(64, 1024, 1536),
(64, 3072, 1024),
(64, 2048, 1536),
@@ -49,7 +47,7 @@ MNK_FACTORS = [
@pytest.mark.parametrize("m,n,k", MNK_FACTORS)
@pytest.mark.parametrize("e", [40, 64, 256])
@pytest.mark.parametrize("topk", [1, 6, 8])
@pytest.mark.parametrize("dtype", [torch.half, torch.bfloat16])
@pytest.mark.parametrize("dtype", [torch.bfloat16])
@torch.inference_mode()
def test_flashinfer_fp4_moe_no_graph(
m: int, n: int, k: int, e: int, topk: int, dtype: torch.dtype