A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 21:30 UTC
[ROCm][CI] Skip DeepGemm-dependent test on ROCm platform (#31462)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Dec 29, 2025 at 01:31 UTC
45c1ca1ca1ee8fa06df263c8715e8a412ff408d4
1 file changed
+4
tests/kernels/moe/test_silu_mul_per_token_group_quant_fp8_colmajor.py
+4
@@ -67,6 +67,10 @@ def reference(x: torch.Tensor, use_ue8m0: bool) -> tuple[torch.Tensor, torch.Ten
67
68
@pytest.mark.parametrize("T", [128, 256, 512])
69
@pytest.mark.parametrize("N", [128 * 2, 256 * 2, 768 * 2, 2048 * 2, 7168 * 2])
70
+@pytest.mark.skipif(
71
+ current_platform.is_rocm(),
72
+ reason="ROCm does not support DeepGemm.",
73
+)
74
def test_silu_mul_fp8_quant_deep_gemm(T: int, N: int):
75
current_platform.seed_everything(42)
76