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

Signed-off-by: shen-shanshan <467638484@qq.com>
2026-02-03 22:47:41 +08:00
parent f3d8a34671
commit 5c4f2dd6ef
15 changed files with 58 additions and 11 deletions
--- a/vllm/model_executor/models/molmo.py
+++ b/vllm/model_executor/models/molmo.py
@@ -231,7 +231,11 @@ class MultiHeadDotProductAttention(nn.Module):

        self.scale = self.head_dim**-0.5
        self.attn = MMEncoderAttention(
-            self.num_heads, self.head_dim, self.scale, num_kv_heads=self.num_kv_heads
+            self.num_heads,
+            self.head_dim,
+            self.scale,
+            num_kv_heads=self.num_kv_heads,
+            prefix=prefix,
        )

    def forward(