trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.608695652173913,
  "eval_steps": 500,
  "global_step": 15,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8695652173913043,
      "grad_norm": 1.6034659147262573,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 3.4258,
      "step": 5
    },
    {
      "epoch": 1.7391304347826086,
      "grad_norm": 1.0319217443466187,
      "learning_rate": 1.2500000000000006e-05,
      "loss": 3.3578,
      "step": 10
    },
    {
      "epoch": 2.608695652173913,
      "grad_norm": 1.2878139019012451,
      "learning_rate": 0.0,
      "loss": 3.2979,
      "step": 15
    },
    {
      "epoch": 2.608695652173913,
      "step": 15,
      "total_flos": 27871774801920.0,
      "train_loss": 3.3604891459147135,
      "train_runtime": 81.3934,
      "train_samples_per_second": 3.354,
      "train_steps_per_second": 0.184
    }
  ],
  "logging_steps": 5,
  "max_steps": 15,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 100,
  "total_flos": 27871774801920.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}