[ROCm] Enable SimpleCPUOffloadConnector on ROCm (#40549)

Signed-off-by: Hongxia Yang <hongxia.yang@amd.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
This commit is contained in:
Hongxia Yang
2026-05-06 20:52:02 -07:00
committed by GitHub
co-authored by TJian
parent 5a0a8fc1ea
commit 20cac26b19
4 changed files with 90 additions and 21 deletions
@@ -10,8 +10,8 @@ from vllm import LLM, SamplingParams, TokensPrompt
from vllm.config import KVTransferConfig
from vllm.platforms import current_platform
if not current_platform.is_cuda():
pytest.skip("Requires CUDA", allow_module_level=True)
if not current_platform.is_cuda_alike():
pytest.skip("Requires CUDA or ROCm", allow_module_level=True)
# Small models for default CI / local runs (accuracy only).
SMALL_MODELS = [