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

[Misc] Clean up cruft from previous FlashMLA sparse implementation (#26125)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
This commit is contained in:
Lucas Wilkinson
2025-10-07 22:09:34 -04:00
committed by GitHub
parent 31a4b3e6c4
commit f80e7866c0
9 changed files with 80 additions and 141 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/kernels/attention/test_attention_selector.py
View File

@@ -165,10 +165,10 @@ def test_env(
pytest.skip("FlashMLA only supports block_size 64")
else:
from vllm.v1.attention.backends.mla.flashmla import (
is_flashmla_supported,
is_flashmla_dense_supported,
)
is_supported, _ = is_flashmla_supported()
is_supported, _ = is_flashmla_dense_supported()
if not is_supported:
pytest.skip("FlashMLA not supported on this platform")
else: