typos

examples/run_glue.py

@@ -116,8 +116,8 @@ def train(args, train_dataset, model, tokenizer):
                       'attention_mask': batch[1],
                       'token_type_ids': batch[2] if args.model_type in ['bert', 'xlnet'] else None,  # XLM don't use segment_ids
                       'labels':         batch[3]}
-            ouputs = model(**inputs)
-            loss = ouputs[0]  # model outputs are always tuple in pytorch-transformers (see doc)
+            outputs = model(**inputs)
+            loss = outputs[0]  # model outputs are always tuple in pytorch-transformers (see doc)
 
             if args.n_gpu > 1:
                 loss = loss.mean() # mean() to average on multi-gpu parallel training