{
  "architecture": "TinyChatLM",
  "model_type": "custom-decoder-only-transformer",
  "vocab_size": 8192,
  "d_model": 192,
  "n_layers": 8,
  "n_heads": 6,
  "n_kv_heads": 2,
  "d_ff": 560,
  "max_seq_len": 512,
  "dropout": 0.0,
  "training": {
    "stage": "sft",
    "step": 3000,
    "val_loss": 3.066343913078308
  }
}