path: root/data/tokenizer_specs.json

{
  "tokenizers": [
    {
      "name": "sp_bpe_1024",
      "dataset_suffix": "sp1024",
      "vocab_size": 1024
    }
  ]
}