tokenizer_config.json

{
  "add_prefix_space": false,
  "bos_token": "<|endoftext|>",
  "eos_token": "<|endoftext|>",
  "name_or_path": "EleutherAI/gpt-neox-20b",
  "tokenizer_class": "GPTNeoXTokenizer",
  "unk_token": "<|endoftext|>", 
  "pad_token": "<|endoftext|>"
}