trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 11.0,
  "eval_steps": 500,
  "global_step": 2200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.5,
      "grad_norm": 2.6582741737365723,
      "learning_rate": 1.5454545454545454e-05,
      "loss": 0.7137,
      "step": 500
    },
    {
      "epoch": 5.0,
      "grad_norm": 11.646175384521484,
      "learning_rate": 1.0909090909090909e-05,
      "loss": 0.7078,
      "step": 1000
    },
    {
      "epoch": 7.5,
      "grad_norm": 3.0001232624053955,
      "learning_rate": 6.363636363636364e-06,
      "loss": 0.7022,
      "step": 1500
    },
    {
      "epoch": 10.0,
      "grad_norm": 5.156288146972656,
      "learning_rate": 1.8181818181818183e-06,
      "loss": 0.6967,
      "step": 2000
    },
    {
      "epoch": 11.0,
      "step": 2200,
      "total_flos": 3.697260574382285e+16,
      "train_loss": 0.703608925559304,
      "train_runtime": 3543.902,
      "train_samples_per_second": 2.48,
      "train_steps_per_second": 0.621
    }
  ],
  "logging_steps": 500,
  "max_steps": 2200,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 11,
  "save_steps": 500,
  "total_flos": 3.697260574382285e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}