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 14, 05:30 UTC
[CI Failure] Skip failing nvfp4 silu test (#23959)
Signed-off-by: mgoin <mgoin64@gmail.com> Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Michael Goin committed
Sep 2, 2025 at 13:18 UTC
e66ed3e675d9e3c8cbc128adf81b62162827daee
1 file changed
+2
-1
tests/kernels/quantization/test_silu_nvfp4_quant_fusion.py
+2
-1
@@ -8,7 +8,8 @@ from vllm.model_executor.layers.activation import SiluAndMul
8
from vllm.platforms import current_platform
9
from vllm.scalar_type import scalar_types
10
11
-if not current_platform.has_device_capability(100):
11
+if not (current_platform.has_device_capability(100)
12
+ and hasattr(torch.ops._C, "silu_and_mul_nvfp4_quant")):
13
pytest.skip(reason="Nvfp4 Requires compute capability of 10 or above.",
14
allow_module_level=True)
15