{
  "tag": "v3",
  "arch": "v2",
  "recipe": "bert",
  "n_layers": 12,
  "n_heads": 12,
  "lr": 0.0001,
  "mode": "autoregressive",
  "epoch": 13,
  "train_loss": 0.97961,
  "val_loss": 4.47436,
  "epoch_sec": 1166.5
}