trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 100,
  "global_step": 80,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.5,
      "learning_rate": 0.000175,
      "loss": 1.2435,
      "step": 10
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.00015000000000000001,
      "loss": 0.7842,
      "step": 20
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.000125,
      "loss": 0.3928,
      "step": 30
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0001,
      "loss": 0.2174,
      "step": 40
    },
    {
      "epoch": 2.5,
      "learning_rate": 7.500000000000001e-05,
      "loss": 0.1137,
      "step": 50
    },
    {
      "epoch": 3.0,
      "learning_rate": 5e-05,
      "loss": 0.0931,
      "step": 60
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.5e-05,
      "loss": 0.0725,
      "step": 70
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0,
      "loss": 0.0704,
      "step": 80
    },
    {
      "epoch": 4.0,
      "step": 80,
      "total_flos": 9.547184254235443e+16,
      "train_loss": 0.37343859672546387,
      "train_runtime": 76.2656,
      "train_samples_per_second": 16.154,
      "train_steps_per_second": 1.049
    }
  ],
  "logging_steps": 10,
  "max_steps": 80,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 100,
  "total_flos": 9.547184254235443e+16,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}