Tokenizer for causal language modeling of DNA sequences

    "vocab": {
      "[PAD]": 0,
      "[UNK]": 1,
      "a": 2,
      "c": 3,
      "g": 4,
      "t": 5,
    },