trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 3100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.61,
      "learning_rate": 4.1935483870967746e-05,
      "loss": 5.3957,
      "step": 500
    },
    {
      "epoch": 3.23,
      "learning_rate": 3.387096774193548e-05,
      "loss": 3.0751,
      "step": 1000
    },
    {
      "epoch": 4.84,
      "learning_rate": 2.5806451612903226e-05,
      "loss": 2.7645,
      "step": 1500
    },
    {
      "epoch": 6.45,
      "learning_rate": 1.774193548387097e-05,
      "loss": 2.5892,
      "step": 2000
    },
    {
      "epoch": 8.06,
      "learning_rate": 9.67741935483871e-06,
      "loss": 2.4902,
      "step": 2500
    },
    {
      "epoch": 9.68,
      "learning_rate": 1.6129032258064516e-06,
      "loss": 2.4229,
      "step": 3000
    },
    {
      "epoch": 10.0,
      "step": 3100,
      "total_flos": 1.677608124908544e+16,
      "train_loss": 3.1009698683215725,
      "train_runtime": 1141.4911,
      "train_samples_per_second": 27.14,
      "train_steps_per_second": 2.716
    }
  ],
  "max_steps": 3100,
  "num_train_epochs": 10,
  "total_flos": 1.677608124908544e+16,
  "trial_name": null,
  "trial_params": null
}