run-14/checkpoint-500/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 6.25,
  "eval_steps": 500,
  "global_step": 500,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.0064516129032258064,
      "eval_loss": NaN,
      "eval_runtime": 1.0637,
      "eval_samples_per_second": 2914.463,
      "eval_steps_per_second": 15.983,
      "step": 80
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.0064516129032258064,
      "eval_loss": NaN,
      "eval_runtime": 1.053,
      "eval_samples_per_second": 2944.004,
      "eval_steps_per_second": 16.145,
      "step": 160
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.0064516129032258064,
      "eval_loss": NaN,
      "eval_runtime": 1.0897,
      "eval_samples_per_second": 2844.817,
      "eval_steps_per_second": 15.601,
      "step": 240
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.0064516129032258064,
      "eval_loss": NaN,
      "eval_runtime": 1.0823,
      "eval_samples_per_second": 2864.215,
      "eval_steps_per_second": 15.707,
      "step": 320
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.0064516129032258064,
      "eval_loss": NaN,
      "eval_runtime": 1.0678,
      "eval_samples_per_second": 2903.113,
      "eval_steps_per_second": 15.92,
      "step": 400
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.0064516129032258064,
      "eval_loss": NaN,
      "eval_runtime": 1.0764,
      "eval_samples_per_second": 2879.866,
      "eval_steps_per_second": 15.793,
      "step": 480
    },
    {
      "epoch": 6.25,
      "learning_rate": 2.1428571428571427e-06,
      "loss": 0.0,
      "step": 500
    }
  ],
  "logging_steps": 500,
  "max_steps": 560,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 7,
  "save_steps": 500,
  "total_flos": 591309336775836.0,
  "train_batch_size": 192,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7974217563133699,
    "num_train_epochs": 7,
    "temperature": 7
  }
}