debugging

modeling_bert.py

@@ -501,6 +501,9 @@ class BertForPreTraining(BertPreTrainedModel):
                 )
         prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output)
 
+        print(labels)
+        assert False
+
         if (
             self.dense_seq_output and labels is not None
         ):  # prediction_scores are already flattened
@@ -515,7 +518,6 @@ class BertForPreTraining(BertPreTrainedModel):
         else:
             masked_lm_loss = 0
         if next_sentence_label is not None:
-            assert False, "TODO: remove this assertion"
             next_sentence_loss = self.nsp_loss(
                 rearrange(seq_relationship_score, "... t -> (...) t"),
                 rearrange(next_sentence_label, "... -> (...)"),