{
  "args": {
    "name": "tinystories-d256-l8-h8-ctx128-5k",
    "batch_size": 32,
    "epochs": 10,
    "max_steps_per_epoch": 500,
    "lr": 0.001,
    "weight_decay": 0.01,
    "seed": 0,
    "run_dir": "runs",
    "eval_prompt": "Once upon a time"
  },
  "model": {
    "d_model": 256,
    "debug": false,
    "layer_norm_eps": 1e-05,
    "d_vocab": 50257,
    "init_range": 0.02,
    "n_ctx": 128,
    "d_head": 32,
    "d_mlp": 1024,
    "n_heads": 8,
    "n_layers": 8
  }
}