trainer_state.json

{
  "best_metric": 0.865,
  "best_model_checkpoint": "final-run/checkpoint-32",
  "epoch": 3.9298245614035086,
  "eval_steps": 500,
  "global_step": 32,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.98,
      "eval_accuracy": 0.71,
      "eval_loss": 1.295009970664978,
      "eval_runtime": 2.2823,
      "eval_samples_per_second": 87.632,
      "eval_steps_per_second": 3.067,
      "step": 8
    },
    {
      "epoch": 1.23,
      "grad_norm": 2.336587429046631,
      "learning_rate": 7.857142857142858e-06,
      "loss": 1.383,
      "step": 10
    },
    {
      "epoch": 1.96,
      "eval_accuracy": 0.815,
      "eval_loss": 1.1531453132629395,
      "eval_runtime": 2.2842,
      "eval_samples_per_second": 87.558,
      "eval_steps_per_second": 3.065,
      "step": 16
    },
    {
      "epoch": 2.46,
      "grad_norm": 2.422740936279297,
      "learning_rate": 4.2857142857142855e-06,
      "loss": 1.2243,
      "step": 20
    },
    {
      "epoch": 2.95,
      "eval_accuracy": 0.86,
      "eval_loss": 1.0636448860168457,
      "eval_runtime": 2.3294,
      "eval_samples_per_second": 85.86,
      "eval_steps_per_second": 3.005,
      "step": 24
    },
    {
      "epoch": 3.68,
      "grad_norm": 2.130319356918335,
      "learning_rate": 7.142857142857143e-07,
      "loss": 1.1189,
      "step": 30
    },
    {
      "epoch": 3.93,
      "eval_accuracy": 0.865,
      "eval_loss": 1.028597354888916,
      "eval_runtime": 2.2993,
      "eval_samples_per_second": 86.984,
      "eval_steps_per_second": 3.044,
      "step": 32
    },
    {
      "epoch": 3.93,
      "step": 32,
      "total_flos": 5.498979435916001e+17,
      "train_loss": 1.2344711497426033,
      "train_runtime": 175.6186,
      "train_samples_per_second": 40.998,
      "train_steps_per_second": 0.182
    }
  ],
  "logging_steps": 10,
  "max_steps": 32,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 5.498979435916001e+17,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}