config.json

{
    "n_layers": 8,
    "d_model": 512,
    "d_mlp": 2048,
    "d_head": 64,
    "n_heads": 8,
    "n_ctx": 59,
    "d_vocab": 61,
    "act_fn": "gelu",
    "attn_only": false,
    "normalization_type": "LN",
    "architecture": "mingpt"
}