Merge pull request #141 from SinghJasdeep/patch-1

loading saved model when n_classes != 2

examples/run_classifier.py

@@ -558,7 +558,7 @@ def main():
 
     # Load a trained model that you have fine-tuned
     model_state_dict = torch.load(output_model_file)
-    model = BertForSequenceClassification.from_pretrained(args.bert_model, state_dict=model_state_dict)
+    model = BertForSequenceClassification.from_pretrained(args.bert_model, state_dict=model_state_dict, num_labels=num_labels)
     model.to(device)
 
     if args.do_eval and (args.local_rank == -1 or torch.distributed.get_rank() == 0):