Update bert_layers.py

bert_layers.py

@@ -495,7 +495,7 @@ class BertEncoder(nn.Module):
             all_attention_weights.append(attention_weights)  # JAANDOUI: appending the attention of different layers together.
             # print(f'here is the matrix of attentions inside encoder: \n {all_attention_weights}')
             # print(f'and this is the shape inside encoder: \n {all_attention_weights.shape}')
-            print(f'NUMBER6: {all_attention_weights}')
+            # print(f'NUMBER6: {all_attention_weights}')
         if not output_all_encoded_layers:
             all_encoder_layers.append(hidden_states)
 
@@ -632,7 +632,7 @@ class BertModel(BertPreTrainedModel):
             attention_mask,
             output_all_encoded_layers=output_all_encoded_layers,
             subset_mask=subset_mask)
-        print(f'NUMBER7: {all_attention_weights}')
+        # print(f'NUMBER7: {all_attention_weights}')
         # print(f'here is the matrix of attentions in BERT: \n {all_attention_weights}')
         # print(f'and this is the shape in BERT: \n {all_attention_weights.shape}')
 
@@ -658,11 +658,11 @@ class BertModel(BertPreTrainedModel):
 
         # JAANDOUI: returning all_attention_weights too
         if self.pooler is not None:
-            print(f'NUMBER8: {all_attention_weights}')
+            # print(f'NUMBER8: {all_attention_weights}')
             return encoder_outputs, pooled_output, all_attention_weights
 
         # JAANDOUI: returning all_attention_weights too
-        print(f'NUMBER9: {all_attention_weights}')
+        # print(f'NUMBER9: {all_attention_weights}')
         return encoder_outputs, None, all_attention_weights
         # JAANDOUI: need to handle the returned elements wherever BertModel is instantiated.
 
@@ -903,7 +903,7 @@ class BertForSequenceClassification(BertPreTrainedModel):
         
         # JAANDOUI:
         all_attention_weights = outputs[2]
-        print(f'last: {all_attention_weights}')
+        # print(f'last: {all_attention_weights}')
 
         pooled_output = self.dropout(pooled_output)
         logits = self.classifier(pooled_output)