best/model_config.json

{
  "vocab_size": 16000,
  "n_layers": 8,
  "n_heads": 8,
  "n_kv_heads": 4,
  "d_model": 512,
  "d_ff": 1536,
  "context_len": 1024,
  "dropout": 0.0,
  "bias": false,
  "norm_eps": 1e-05,
  "rope_theta": 10000.0,
  "tie_weights": true
}