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

replace cuda_device_count_stateless() to current_platform.device_count() (#37841)

Browse Source 
Signed-off-by: Liao, Wei <wei.liao@intel.com>
Signed-off-by: wliao2 <wei.liao@intel.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
wliao2
2026-03-31 07:32:54 -07:00
committed by GitHub
parent e8057c00bc
commit 4dfad17ed1
20 changed files with 96 additions and 92 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/models/quantization/test_gpt_oss.py
View File

@@ -21,8 +21,8 @@ import lm_eval
import pytest
from packaging import version
from vllm.platforms import current_platform
from vllm.platforms.rocm import on_gfx950
from vllm.utils.torch_utils import cuda_device_count_stateless
MODEL_ACCURACIES = {
# Full quantization: attention linears and MoE linears
@@ -89,7 +89,7 @@ def test_gpt_oss_attention_quantization(
expected_accuracy: float,
monkeypatch: pytest.MonkeyPatch,
):
if tp_size > cuda_device_count_stateless():
if tp_size > current_platform.device_count():
pytest.skip("Not enough GPUs to run this test case")
if "amd/gpt-oss-20b-MoE-Quant-W-MXFP4-A-FP8-KV-FP8" in model_name and on_gfx950():