[MM] Pass prefix parameter to MMEncoderAttention (#33674)

Signed-off-by: shen-shanshan <467638484@qq.com>
This commit is contained in:
Shanshan Shen
2026-02-03 06:47:41 -08:00
committed by GitHub
parent f3d8a34671
commit 5c4f2dd6ef
15 changed files with 58 additions and 11 deletions
+4 -1
View File
@@ -127,7 +127,10 @@ class AIMv2Attention(nn.Module):
self.num_heads_per_partition = divide(self.num_heads, self.tp_size)
self.attn = MMEncoderAttention(
self.num_heads_per_partition, self.head_dim, self.scale
self.num_heads_per_partition,
self.head_dim,
self.scale,
prefix=prefix,
)
def forward(self, x: torch.Tensor) -> torch.Tensor: