trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 3020,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.66,
      "learning_rate": 4.1721854304635764e-05,
      "loss": 2.9027,
      "step": 500
    },
    {
      "epoch": 3.31,
      "learning_rate": 3.3443708609271526e-05,
      "loss": 1.5926,
      "step": 1000
    },
    {
      "epoch": 4.97,
      "learning_rate": 2.5165562913907287e-05,
      "loss": 0.9264,
      "step": 1500
    },
    {
      "epoch": 6.62,
      "learning_rate": 1.688741721854305e-05,
      "loss": 0.4668,
      "step": 2000
    },
    {
      "epoch": 8.28,
      "learning_rate": 8.609271523178809e-06,
      "loss": 0.2535,
      "step": 2500
    },
    {
      "epoch": 9.93,
      "learning_rate": 3.3112582781456954e-07,
      "loss": 0.1434,
      "step": 3000
    },
    {
      "epoch": 10.0,
      "step": 3020,
      "total_flos": 1.2491446031941632e+16,
      "train_loss": 1.0414469499461698,
      "train_runtime": 1587.1892,
      "train_samples_per_second": 18.983,
      "train_steps_per_second": 1.903
    }
  ],
  "max_steps": 3020,
  "num_train_epochs": 10,
  "total_flos": 1.2491446031941632e+16,
  "trial_name": null,
  "trial_params": null
}