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

Remove V0 attention backends (#25351)

Browse Source 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Woosuk Kwon
2025-09-21 16:03:28 -07:00
committed by yewentao256
parent 319966a678
commit a815d820ee
28 changed files with 143 additions and 7376 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/compile/test_fusion_attn.py
View File

@ -334,8 +334,9 @@ else:
[7, 256, 533] if current_platform.is_cuda() else [8])
@pytest.mark.parametrize("dtype", [torch.bfloat16, torch.float16])
@pytest.mark.parametrize("model_name, model_class", MODELS)
@pytest.mark.parametrize("backend", [_Backend.FLASHINFER] if
current_platform.is_cuda() else [_Backend.ROCM_FLASH])
@pytest.mark.parametrize("backend",
[_Backend.FLASHINFER] if current_platform.is_cuda()
else [_Backend.TRITON_ATTN_VLLM_V1])
@pytest.mark.parametrize(
"split_attention",
[False, True] if current_platform.is_rocm() else [False])