Set tokenizer class to `PreTrainedTokenizerFast`

tokenizer_config.json

@@ -8008,6 +8008,6 @@
   "clean_up_tokenization_spaces": false,
   "eos_token": "</s>",
   "model_max_length": 1000000000000000019884624838656,
-  "tokenizer_class": "GPT2Tokenizer",
+  "tokenizer_class": "PreTrainedTokenizerFast",
   "unk_token": "<unk>"
 }