run-1/checkpoint-1000/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.3404825737265416,
  "eval_steps": 500,
  "global_step": 1000,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.6702412868632708,
      "grad_norm": 23.28346824645996,
      "learning_rate": 9.798511296679954e-06,
      "loss": 0.5531,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8025477528572083,
      "eval_loss": 0.4079710841178894,
      "eval_runtime": 9.2612,
      "eval_samples_per_second": 322.097,
      "eval_steps_per_second": 20.192,
      "step": 746
    },
    {
      "epoch": 1.3404825737265416,
      "grad_norm": 38.64032745361328,
      "learning_rate": 9.094595255251797e-06,
      "loss": 0.3957,
      "step": 1000
    }
  ],
  "logging_steps": 500,
  "max_steps": 7460,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 2642653548489780.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 1.0502427338108112e-05,
    "per_device_train_batch_size": 16
  }
}