trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 2994,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.5,
      "grad_norm": 24.274106979370117,
      "learning_rate": 4.1649966599866405e-05,
      "loss": 1.675,
      "step": 500
    },
    {
      "epoch": 1.0,
      "grad_norm": 7.8456525802612305,
      "learning_rate": 3.32999331997328e-05,
      "loss": 1.0539,
      "step": 1000
    },
    {
      "epoch": 1.5,
      "grad_norm": 20.231595993041992,
      "learning_rate": 2.49498997995992e-05,
      "loss": 0.4262,
      "step": 1500
    },
    {
      "epoch": 2.0,
      "grad_norm": 17.46414566040039,
      "learning_rate": 1.65998663994656e-05,
      "loss": 0.3978,
      "step": 2000
    },
    {
      "epoch": 2.51,
      "grad_norm": 1.7042111158370972,
      "learning_rate": 8.249832999331998e-06,
      "loss": 0.1168,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2994,
      "total_flos": 4689746024374080.0,
      "train_loss": 0.6347126753710872,
      "train_runtime": 558.4324,
      "train_samples_per_second": 42.859,
      "train_steps_per_second": 5.361
    }
  ],
  "logging_steps": 500,
  "max_steps": 2994,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 4689746024374080.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}