Update tokenizer_config.json

remove sep_token because it led to an error during inference

tokenizer_config.json

@@ -33,7 +33,6 @@
   "legacy": true,
   "model_max_length": 1000000000000000019884624838656,
   "pad_token": "<pad>",
-  "sep_token": "[SEP]",
   "sp_model_kwargs": {},
   "tokenizer_class": "T5Tokenizer",
   "unk_token": "<unk>"