trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2775,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.54,
      "learning_rate": 4.099099099099099e-05,
      "loss": 1.9648,
      "step": 500
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.198198198198199e-05,
      "loss": 1.2606,
      "step": 1000
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.2972972972972976e-05,
      "loss": 1.0457,
      "step": 1500
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.3963963963963963e-05,
      "loss": 0.9155,
      "step": 2000
    },
    {
      "epoch": 2.7,
      "learning_rate": 4.954954954954955e-06,
      "loss": 0.846,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2775,
      "total_flos": 2.587401364124467e+16,
      "train_loss": 1.167376989416174,
      "train_runtime": 1369.5021,
      "train_samples_per_second": 20.256,
      "train_steps_per_second": 2.026
    }
  ],
  "max_steps": 2775,
  "num_train_epochs": 3,
  "total_flos": 2.587401364124467e+16,
  "trial_name": null,
  "trial_params": null
}