Update tokenization_openai.py

pytorch_transformers/tokenization_openai.py

@@ -92,6 +92,7 @@ class OpenAIGPTTokenizer(PreTrainedTokenizer):
             from spacy.lang.en import English
             _nlp = English()
             self.nlp = nlp.Defaults.create_tokenizer(_nlp)
+            self.fix_text = ftfy.fix_text
         except ImportError:
             logger.warning("ftfy or spacy is not installed using BERT BasicTokenizer instead of SpaCy & ftfy.")
             self.nlp = BasicTokenizer(do_lower_case=True)