Allow for attention weights to be extracted.

modeling_codesage.py

@@ -149,7 +149,7 @@ class CodeSageBlock(nn.Module):
         feed_forward_hidden_states = self.mlp(hidden_states)
         hidden_states = residual + feed_forward_hidden_states
 
-        outputs = (hidden_states,) + outputs[1:]
+        outputs = (hidden_states,) + outputs
         return outputs  # hidden_states, present, (attentions)