trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 7.384615384615385,
  "eval_steps": 500,
  "global_step": 24,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.54,
      "learning_rate": 4.4833833507280884e-05,
      "loss": 2.6796,
      "step": 5
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.147047612756302e-05,
      "loss": 2.5826,
      "step": 10
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.5432914190872757e-05,
      "loss": 2.5076,
      "step": 15
    },
    {
      "epoch": 6.15,
      "learning_rate": 3.3493649053890326e-06,
      "loss": 2.4604,
      "step": 20
    },
    {
      "epoch": 7.38,
      "step": 24,
      "total_flos": 1.6784192100630528e+16,
      "train_loss": 2.546178142229716,
      "train_runtime": 133.3991,
      "train_samples_per_second": 3.118,
      "train_steps_per_second": 0.18
    }
  ],
  "logging_steps": 5,
  "max_steps": 24,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 8,
  "save_steps": 100,
  "total_flos": 1.6784192100630528e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}