tokenizer_config.json

{
  "tokenizer_class": "ByT5Tokenizer",
  "pretrained_model_name_or_path": "google/byt5-small",
  "base_vocab_size": 384,
  "model_embedding_vocab_size": 4096,
  "pad_token": "<pad>",
  "eos_token": "</s>",
  "unk_token": "<unk>",
  "pad_token_id": 0,
  "eos_token_id": 1,
  "unk_token_id": 2,
  "cls_token": "<cls>",
  "cls_token_id": 384,
  "sep_token": "<sep>",
  "sep_token_id": 385,
  "max_length": 128,
  "tokenizer_semantics": "raw UTF-8 byte -> byte + 3"
}