set encoding to 'utf-8' in calls to open

pytorch_pretrained_bert/file_utils.py

@@ -227,7 +227,7 @@ def read_set_from_file(filename: str) -> Set[str]:
     Expected file format is one item per line.
     '''
     collection = set()
-    with open(filename, 'r') as file_:
+    with open(filename, 'r', encoding='utf-8') as file_:
         for line in file_:
             collection.add(line.rstrip())
     return collection