training_summary.json

{
  "model": "amor_gdn",
  "scale": "180m",
  "final_val_loss": 3.4208580493927,
  "final_val_ppl": 30.59565634561051,
  "best_val_loss": 3.4207578182220457,
  "best_val_ppl": 30.592589860839297,
  "total_steps": 7503,
  "tokens_trained": 3687948288,
  "training_hours": 12.336112522151735,
  "n_params": 181967832,
  "flops_per_token": 983531520.0,
  "world_size": 1,
  "linear_flops": 469237927772160,
  "attention_flops": 14187484938240,
  "ssm_flops": 0,
  "final_fire_rate": 0.3932831282663326,
  "final_alpha": 0.8205058963875017
}