trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9984,
  "eval_steps": 50,
  "global_step": 104,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.24,
      "learning_rate": 2.5e-05,
      "loss": 4.1382,
      "step": 25
    },
    {
      "epoch": 0.48,
      "learning_rate": 5e-05,
      "loss": 3.621,
      "step": 50
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.500000000000001e-05,
      "loss": 2.0033,
      "step": 75
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0001,
      "loss": 0.9072,
      "step": 100
    }
  ],
  "logging_steps": 25,
  "max_steps": 1000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 100,
  "total_flos": 7881509745008640.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}