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 8, 02:30 UTC
[CI][AMD][BugFix] Skip tests in test_unquantized_backend_selection that should not run on ROCm (#34655)
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
rasmith committed
Feb 18, 2026 at 14:00 UTC
fcd6ac97ed0c26b36dbf18a1ee8ecb5f6b41707e
1 file changed
+7
tests/kernels/moe/test_unquantized_backend_selection.py
+7
@@ -9,6 +9,7 @@ from vllm.model_executor.layers.fused_moe.oracle.unquantized import (
9
UnquantizedMoeBackend,
10
select_unquantized_moe_backend,
11
)
12
+from vllm.platforms import current_platform
13
14
15
@pytest.mark.parametrize(
@@ -65,6 +66,9 @@ def test_select_default_backend_by_platform(
66
"vllm.model_executor.layers.fused_moe.oracle.unquantized.is_supported_config_trtllm_bf16",
67
return_value=(True, None),
68
)
69
+@pytest.mark.skipif(
70
+ not current_platform.is_cuda(), reason="Only supported on NVIDIA platforms."
71
+)
72
def test_select_cuda_flashinfer_trtllm_backend(
73
mock_has_flashinfer, mock_is_supported_trtllm, monkeypatch
74
):
@@ -101,6 +105,9 @@ def test_select_cuda_flashinfer_trtllm_backend(
105
"vllm.model_executor.layers.fused_moe.oracle.unquantized.is_supported_config_trtllm_bf16",
106
return_value=(False, None),
107
)
108
+@pytest.mark.skipif(
109
+ not current_platform.is_cuda(), reason="Only supported on NVIDIA platforms."
110
+)
111
def test_select_cuda_flashinfer_cutlass_backend(
112
mock_has_flashinfer, mock_is_supported_trtllm, monkeypatch
113
):