weights/tokenizer.json

{
  "pad_id": 0,
  "bos_id": 1,
  "sep_id": 2,
  "caps_len": 32,
  "type": "char+byte-fallback-buckets",
  "charset": "abcdefghijklmnopqrstuvwxyz ,:-:'",
  "note": "ids: 0 pad,1 bos,2 sep,3-6 punct,7-32 a-z,33-62 byte buckets"
}