trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1761,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.85,
      "learning_rate": 3.580352072685974e-05,
      "loss": 6.4049,
      "step": 500
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.160704145371948e-05,
      "loss": 5.7387,
      "step": 1000
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.410562180579217e-06,
      "loss": 5.5475,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1761,
      "total_flos": 3677946642432000.0,
      "train_loss": 5.833661626916525,
      "train_runtime": 1259.5482,
      "train_samples_per_second": 5.588,
      "train_steps_per_second": 1.398
    }
  ],
  "logging_steps": 500,
  "max_steps": 1761,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 3677946642432000.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}