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, 04:15 UTC
[CI/Build] Skip cpu offloading test on AMD (#27690)
Signed-off-by: zhewenli <zhewenli@meta.com>
Zhewen Li committed
Oct 29, 2025 at 05:55 UTC
9a0d2f0d92908c7fcc17eab63a3dfc998aae53a4
1 file changed
+4
tests/v1/kv_offload/test_cpu_offloading.py
+4
@@ -12,6 +12,7 @@ from tqdm import tqdm
12
from vllm import LLM, SamplingParams, TokensPrompt
13
from vllm.config import KVEventsConfig, KVTransferConfig
14
from vllm.distributed.kv_events import BlockStored, KVEventBatch
15
+from vllm.platforms import current_platform
16
17
CPU_BLOCK_SIZES = [16, 48]
18
@@ -63,6 +64,9 @@ class MockSubscriber:
64
self.sub.close()
65
66
67
+@pytest.mark.skipif(
68
+ not current_platform.is_cuda(), reason="CPU offloading only supported on CUDA"
69
+)
70
@pytest.mark.parametrize("cpu_block_size", CPU_BLOCK_SIZES)
71
def test_cpu_offloading(cpu_block_size: int) -> None:
72
"""