frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

test_attention compat with coming xformers change (#20487)

Browse Source 
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Jeremy Reizenstein
2025-07-06 03:37:59 +01:00
committed by GitHub
parent 8d763cb891
commit c5ebe040ac
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/kernels/attention/test_attention.py
View File

@ -450,7 +450,8 @@ def test_multi_query_kv_attention(
start += seq_len
# xformers.AttentionBias to Tensor for use in reference impl.
alibi_bias = [
b.materialize(b.shape, device=device).squeeze() for b in attn_bias
b.materialize((1, num_query_heads, i, i), device=device).squeeze()
for b, i in zip(attn_bias, seq_lens)
]
else:
attn_bias = BlockDiagonalCausalMask.from_seqlens(seq_lens)