switch to the API in torchtext 0.9.0

examples/BERT/cross_lingual_mlm_task.py

@@ -138,7 +138,7 @@ def text_transform(x: str) -> List:
 
     for epoch in range(1, args.epochs + 1):
         train_data = CC100('/datasets01/cc100/031720/', {'*.txt'}, start_line=args.start_line, num_lines=args.num_lines)
-        from torchtext.experimental.datasets.raw import WikiText2
+        from torchtext.datasets import WikiText2
         val_data, = WikiText2(data_select='valid')
         val_data = [(17, item) for item in val_data if item != ' \n']  # english language type is 17 in CC100 dataset