training_summary.json

{
  "model": "amor_gdn",
  "scale": "1.5b",
  "final_val_loss": 2.572668752670288,
  "final_val_ppl": 13.100740465193947,
  "best_val_loss": 2.5728520250320432,
  "best_val_ppl": 13.103141688871991,
  "total_steps": 62558,
  "tokens_trained": 30748520448,
  "training_hours": 9.102619256443447,
  "n_params": 1524025472,
  "flops_per_token": 4699619328.0,
  "world_size": 1,
  "linear_flops": 2272123598929920,
  "attention_flops": 37833293168640,
  "ssm_flops": 0,
  "final_fire_rate": 0.4176534337057724,
  "final_alpha": 0.507554063081953
}