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 5, 04:45 UTC
[ROCm][CI] Fix condition for `test_per_token_group_quant_fp8_packed` (#39730)
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
Micah Williamson committed
Apr 14, 2026 at 11:14 UTC
4f4f8eaa78fb861dcb0b141951f7496b47ea679e
1 file changed
+4
-1
tests/kernels/quantization/test_per_token_group_quant.py
+4
-1
@@ -6,6 +6,7 @@ import pytest
6
import torch
7
8
from vllm.model_executor.layers.quantization.utils import fp8_utils, int8_utils
9
+from vllm.platforms import current_platform
10
11
12
@pytest.mark.parametrize(
@@ -80,7 +81,9 @@ def test_per_token_group_quant_fp8(
81
],
82
)
83
@pytest.mark.parametrize("poisoned_scales", [False, True])
83
-@pytest.mark.skipif(not torch.cuda.is_available(), reason="CUDA not available")
84
+@pytest.mark.skipif(
85
+ not current_platform.is_cuda(), reason="DeepGEMM not available on this platform"
86
+)
87
def test_per_token_group_quant_fp8_packed(
88
num_tokens, hidden_dim, group_size, poisoned_scales
89
):