Upload modeling_opt.py

modeling_opt.py

@@ -184,7 +184,7 @@ class OPTAttention(nn.Module):
         attn_gate_mlp2=False,
         attn_gate_linear_all_features=False,
         fine_tuning=False,
-        attn_softmax=softmax,
+        attn_softmax='vanilla',
     ):
         super().__init__()
         self.embed_dim = embed_dim