trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 50.0,
  "eval_steps": 500,
  "global_step": 2350,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 10.64,
      "learning_rate": 3.936170212765958e-05,
      "loss": 0.7712,
      "step": 500
    },
    {
      "epoch": 21.28,
      "learning_rate": 2.8723404255319154e-05,
      "loss": 0.0462,
      "step": 1000
    },
    {
      "epoch": 31.91,
      "learning_rate": 1.8085106382978724e-05,
      "loss": 0.0141,
      "step": 1500
    },
    {
      "epoch": 42.55,
      "learning_rate": 7.446808510638298e-06,
      "loss": 0.005,
      "step": 2000
    },
    {
      "epoch": 50.0,
      "step": 2350,
      "total_flos": 2.93958851328e+16,
      "train_loss": 0.1785320406264447,
      "train_runtime": 1042.289,
      "train_samples_per_second": 143.914,
      "train_steps_per_second": 2.255
    }
  ],
  "logging_steps": 500,
  "max_steps": 2350,
  "num_train_epochs": 50,
  "save_steps": 500,
  "total_flos": 2.93958851328e+16,
  "trial_name": null,
  "trial_params": null
}