trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 2500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 12.050017356872559,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.653,
      "step": 500
    },
    {
      "epoch": 2.0,
      "grad_norm": 15.045541763305664,
      "learning_rate": 1.2e-05,
      "loss": 0.4015,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "grad_norm": 11.139913558959961,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.2259,
      "step": 1500
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.034658435732126236,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0796,
      "step": 2000
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.020124856382608414,
      "learning_rate": 0.0,
      "loss": 0.0347,
      "step": 2500
    },
    {
      "epoch": 5.0,
      "step": 2500,
      "total_flos": 4659656816640000.0,
      "train_loss": 0.27893603820800783,
      "train_runtime": 3930.0353,
      "train_samples_per_second": 5.089,
      "train_steps_per_second": 0.636
    }
  ],
  "logging_steps": 500,
  "max_steps": 2500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 4659656816640000.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}