{
  "best_metric": 0.7906066536203522,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-16/checkpoint-110",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 165,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.115179181098938,
      "learning_rate": 0.0001711445020920055,
      "loss": 0.4773,
      "step": 55
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7181996086105675,
      "eval_f1": 0.7125748502994013,
      "eval_loss": 0.4258935749530792,
      "eval_precision": 0.7270875763747454,
      "eval_recall": 0.6986301369863014,
      "eval_runtime": 27.8796,
      "eval_samples_per_second": 36.658,
      "eval_steps_per_second": 1.148,
      "step": 55
    },
    {
      "epoch": 2.0,
      "grad_norm": 2.2601873874664307,
      "learning_rate": 0.0001283583765690041,
      "loss": 0.4242,
      "step": 110
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7906066536203522,
      "eval_f1": 0.8018518518518519,
      "eval_loss": 0.40127235651016235,
      "eval_precision": 0.7609841827768014,
      "eval_recall": 0.8473581213307241,
      "eval_runtime": 27.9127,
      "eval_samples_per_second": 36.614,
      "eval_steps_per_second": 1.146,
      "step": 110
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.3176016807556152,
      "learning_rate": 8.557225104600275e-05,
      "loss": 0.4012,
      "step": 165
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7504892367906066,
      "eval_f1": 0.7362978283350569,
      "eval_loss": 0.38755813241004944,
      "eval_precision": 0.7807017543859649,
      "eval_recall": 0.6966731898238747,
      "eval_runtime": 28.6354,
      "eval_samples_per_second": 35.69,
      "eval_steps_per_second": 1.117,
      "step": 165
    }
  ],
  "logging_steps": 500,
  "max_steps": 275,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 56,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.6575195229183988,
    "learning_rate": 0.00021393062761500687,
    "num_train_epochs": 5,
    "per_device_train_batch_size": 56,
    "temperature": 22
  }
}