trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 2166,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.69,
      "learning_rate": 3.845798707294552e-05,
      "loss": 1.029,
      "step": 500
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.6915974145891044e-05,
      "loss": 0.8087,
      "step": 1000
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5373961218836565e-05,
      "loss": 0.7124,
      "step": 1500
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.831948291782087e-06,
      "loss": 0.5622,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2166,
      "total_flos": 2284847800184592.0,
      "train_loss": 0.7604470539181076,
      "train_runtime": 647.9763,
      "train_samples_per_second": 106.856,
      "train_steps_per_second": 3.343
    }
  ],
  "logging_steps": 500,
  "max_steps": 2166,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 2284847800184592.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}