[MM] Align the prefix of MMEncoderAttention with Attention (#33750)

Signed-off-by: shen-shanshan <467638484@qq.com>
2026-02-04 12:07:30 +08:00
parent 4dffc5e044
commit 9fb27dd3b3
17 changed files with 17 additions and 15 deletions
--- a/vllm/model_executor/models/molmo.py
+++ b/vllm/model_executor/models/molmo.py
@@ -235,7 +235,7 @@ class MultiHeadDotProductAttention(nn.Module):
            self.head_dim,
            self.scale,
            num_kv_heads=self.num_kv_heads,
-            prefix=prefix,
+            prefix=f"{prefix}.attn",
        )

    def forward(