[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
+1
View File
@@ -296,6 +296,7 @@ class Glm4vVisionAttention(nn.Module):
num_heads=self.num_attention_heads_per_partition,
head_size=self.hidden_size_per_attention_head,
scale=self.hidden_size_per_attention_head**-0.5,
prefix=prefix,
)
self.apply_rotary_emb = ApplyRotaryEmb(enforce_enable=True)