{
  "model_type": "llama2.c",
  "base_repo": "karpathy/llama2.c",
  "dim": 128,
  "n_layers": 4,
  "n_heads": 4,
  "vocab_size": 2048,
  "max_seq_len": 64
}