{
  "vocab_size": 16384,
  "context_length": 128,
  "d_model": 256,
  "n_heads": 4,
  "n_layers": 4,
  "d_ff": 1024,
  "dropout": 0.0
}