training_summary.json

{
  "model": "amor_gdn",
  "scale": "440m",
  "final_val_loss": 2.958676857948303,
  "final_val_ppl": 19.272454682639406,
  "best_val_loss": 2.9591263580322265,
  "best_val_ppl": 19.281119599931287,
  "total_steps": 18255,
  "tokens_trained": 8973017088,
  "training_hours": 21.794289468195704,
  "n_params": 442133056,
  "flops_per_token": 1710047232.0,
  "world_size": 1,
  "linear_flops": 821605768888320,
  "attention_flops": 18916646584320,
  "ssm_flops": 0,
  "final_fire_rate": 0.4086735447457172,
  "final_alpha": 0.5992211346782192
}