[Bugfix] Only pad transformers backend value when it is narrower (#50060)

Signed-off-by: Nick Hill <nickhill123@gmail.com>
This commit is contained in:
Nick Hill
2026-07-27 23:36:45 -05:00
committed by GitHub
parent 52c3c4a42f
commit d223c900d8
@@ -64,12 +64,15 @@ def vllm_attention_forward(
head_dim_v = value.shape[-1]
query, key, value = (x.transpose(1, 2) for x in (query, key, value))
query, key, value = (x.reshape(hidden, -1) for x in (query, key, value))
# Pad `value` up to the query/key head size when they differ (expanded MLA).
if head_dim_v != head_dim_qk:
# Pad `value` up to the query/key head size when it is smaller (expanded
# MLA). A larger last dim just means `value` isn't split per head, e.g.
# packed grouped/multi-query projections, and needs no padding.
pad_value = head_dim_v < head_dim_qk
if pad_value:
value = F.pad(value.view(-1, head_dim_v), (0, head_dim_qk - head_dim_v))
value = value.reshape(hidden, -1)
attn_output = self_attn.forward(query, key, value)
if head_dim_v != head_dim_qk:
if pad_value:
attn_output = attn_output.view(-1, head_dim_qk)[..., :head_dim_v]
attn_output = attn_output.reshape(hidden, -1)
return attn_output, None