trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1176,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.28,
      "grad_norm": 25.054931640625,
      "learning_rate": 2.8741496598639456e-05,
      "loss": 2.6571,
      "step": 500
    },
    {
      "epoch": 2.55,
      "grad_norm": 19.457901000976562,
      "learning_rate": 7.482993197278912e-06,
      "loss": 1.7886,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1176,
      "total_flos": 4376310425967942.0,
      "train_loss": 2.0936863049357926,
      "train_runtime": 1316.5495,
      "train_samples_per_second": 57.147,
      "train_steps_per_second": 0.893
    }
  ],
  "logging_steps": 500,
  "max_steps": 1176,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 4376310425967942.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}