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

refactor hard coded device string in test files under tests/v1 and tests/lora (#37566)

Browse Source 
Signed-off-by: Liao, Wei <wei.liao@intel.com>
This commit is contained in:
wliao2
2026-04-02 20:21:47 -07:00
committed by GitHub
parent 4a06e1246e
commit 32e0c0bfa2
28 changed files with 239 additions and 146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/v1/spec_decode/test_tree_attention.py
View File

@@ -18,6 +18,8 @@ from vllm.v1.attention.backend import CommonAttentionMetadata
from vllm.v1.attention.backends.fa_utils import is_flash_attn_varlen_func_available
from vllm.v1.attention.backends.registry import AttentionBackendEnum
DEVICE_TYPE = current_platform.device_type
if not is_flash_attn_varlen_func_available():
pytest.skip(
"This test requires flash_attn_varlen_func, but it's not available.",
@@ -170,9 +172,9 @@ def _get_available_reference_backends() -> list[AttentionBackendEnum]:
class MockAttentionLayer(torch.nn.Module):
_q_scale = torch.tensor(1.0, dtype=torch.float32, device="cuda")
_k_scale = torch.tensor(1.0, dtype=torch.float32, device="cuda")
_v_scale = torch.tensor(1.0, dtype=torch.float32, device="cuda")
_q_scale = torch.tensor(1.0, dtype=torch.float32, device=DEVICE_TYPE)
_k_scale = torch.tensor(1.0, dtype=torch.float32, device=DEVICE_TYPE)
_v_scale = torch.tensor(1.0, dtype=torch.float32, device=DEVICE_TYPE)
layer_name = "mock_layer"
def __init__(self):