trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 2100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.38,
      "learning_rate": 3.809523809523809e-05,
      "loss": 4.9012,
      "step": 500
    },
    {
      "epoch": 4.76,
      "learning_rate": 2.6190476190476192e-05,
      "loss": 1.5533,
      "step": 1000
    },
    {
      "epoch": 7.14,
      "learning_rate": 1.4285714285714285e-05,
      "loss": 1.3399,
      "step": 1500
    },
    {
      "epoch": 9.52,
      "learning_rate": 2.3809523809523808e-06,
      "loss": 0.5929,
      "step": 2000
    },
    {
      "epoch": 10.0,
      "step": 2100,
      "total_flos": 6576709761269760.0,
      "train_loss": 2.022017553420294,
      "train_runtime": 872.0967,
      "train_samples_per_second": 24.08,
      "train_steps_per_second": 2.408
    }
  ],
  "max_steps": 2100,
  "num_train_epochs": 10,
  "total_flos": 6576709761269760.0,
  "trial_name": null,
  "trial_params": null
}