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

[ROCm] Skip tests for quantizations incompatible with ROCm (#17905)

Browse Source 
Signed-off-by: Hissu Hyvarinen <hissu.hyvarinen@amd.com>
This commit is contained in:
hissu-hyvarinen
2025-05-13 03:39:28 +03:00
committed by GitHub
parent d67085c2c8
commit f6518b2b48
4 changed files with 20 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/models/quantization/test_gptq_marlin_24.py
View File

@@ -10,6 +10,7 @@ from dataclasses import dataclass
import pytest
from tests.quantization.utils import is_quant_method_supported
from vllm.platforms import current_platform
from ..utils import check_logprobs_close
@@ -38,7 +39,9 @@ model_pairs = [
@pytest.mark.flaky(reruns=2)
@pytest.mark.skipif(not is_quant_method_supported("gptq_marlin_24"),
@pytest.mark.skipif(not is_quant_method_supported("gptq_marlin_24")
or current_platform.is_rocm()
or not current_platform.is_cuda(),
reason="Marlin24 is not supported on this GPU type.")
@pytest.mark.parametrize("model_pair", model_pairs)
@pytest.mark.parametrize("dtype", ["half"])