Update bert_layers.py

bert_layers.py

@@ -954,10 +954,11 @@ class BertForSequenceClassification(BertPreTrainedModel):
             print(f'not stacked final attention SHAPE: {outputs[2].shape}')
         except:
             print(f'not stacked final attention LEN: {len(outputs[2])}')
-                try:
+        try:
             print(f'STACKED final attention SHAPE: {(torch.stack(outputs[2], dim=0)).shape}')
         except:
             print(f'STACKED final attention LEN: {len(torch.stack(outputs[2], dim=0))}')
+            
         return SequenceClassifierOutput(
             loss=loss,
             logits=logits,