Update modeling_bert.py

modeling_bert.py

@@ -286,7 +286,7 @@ class BertSelfAttention(nn.Module):
         self,
         config,
         position_embedding_type=None,
-        softmax_fn=partial(clipped_softmax, gamma=-0.025, eta=1.0),
+        softmax_fn=partial(clipped_softmax1, gamma=-0.025, eta=1.0),
         max_seq_length = 128,
         alpha=None,
         ssm_eps=None,