trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1857,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8077544426494345,
      "grad_norm": 2.0110491628996834,
      "learning_rate": 3.653742595584276e-05,
      "loss": 2.576,
      "step": 500
    },
    {
      "epoch": 1.615508885298869,
      "grad_norm": 1.922235177894796,
      "learning_rate": 2.3074851911685516e-05,
      "loss": 1.9215,
      "step": 1000
    },
    {
      "epoch": 2.4232633279483036,
      "grad_norm": 2.3049251003554425,
      "learning_rate": 9.612277867528271e-06,
      "loss": 1.4342,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1857,
      "total_flos": 5.480111913004171e+18,
      "train_loss": 1.8135224785545403,
      "train_runtime": 5570.2565,
      "train_samples_per_second": 21.323,
      "train_steps_per_second": 0.333
    }
  ],
  "logging_steps": 500,
  "max_steps": 1857,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 5.480111913004171e+18,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}