A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 20:15 UTC
[Bug] Fix log issue with `\n` (#31390)
Signed-off-by: yewentao256 <zhyanwentao@126.com> Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Wentao Ye committed
Dec 31, 2025 at 00:16 UTC
357d435c549dbfcce9f1dd2c737750d7046cffee
1 file changed
+2
-2
vllm/v1/attention/backends/mla/flashmla_sparse.py
+2
-2
@@ -930,8 +930,8 @@ class FlashMLASparseImpl(MLACommonBaseImpl[FlashMLASparseMetadata]):
930
if self.num_heads % self.padding != 0:
931
assert self.padding % self.num_heads == 0
932
logger.warning_once(
933
- f"padding num_heads to {self.padding} \
934
- due to sparse attn kernel requirement"
933
+ f"padding num_heads to {self.padding} due to sparse attn "
934
+ "kernel requirement"
935
)
936
q_padded = q.new_empty((q.shape[0], self.padding, q.shape[2]))
937
q_padded[:, : self.num_heads, :] = q