step_220000/config.json

{
  "d_model": 128,
  "nhead": 4,
  "num_layers": 4,
  "vocab_size": 257
}