Patch tokenizer_class to PreTrainedTokenizerFast

tokenizer_config.json

@@ -59,6 +59,6 @@
   "mask_token": "<mask>",
   "model_max_length": 1000000000000000019884624838656,
   "pad_token": "<pad>",
-  "tokenizer_class": "BertProteinSequenceTokenizer",
+  "tokenizer_class": "PreTrainedTokenizerFast",
   "unk_token": "<unk>"
-}
+}