run-4/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": 13.804513931274414,
      "learning_rate": 1.2284160000241888e-05,
      "loss": 0.5698,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7532685399055481,
      "eval_loss": 0.4876743257045746,
      "eval_runtime": 9.5369,
      "eval_samples_per_second": 312.785,
      "eval_steps_per_second": 19.608,
      "step": 746
    },
    {
      "epoch": 1.3404825737265416,
      "grad_norm": 23.23216438293457,
      "learning_rate": 1.1401677241603822e-05,
      "loss": 0.4247,
      "step": 1000
    }
  ],
  "logging_steps": 500,
  "max_steps": 7460,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 2602962871446816.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 1.3166642758879955e-05,
    "per_device_train_batch_size": 16
  }
}