{
  "best_metric": 0.7504892367906066,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-21/checkpoint-104",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 156,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 2.939810276031494,
      "learning_rate": 0.00013471680923458288,
      "loss": 0.5033,
      "step": 52
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7152641878669276,
      "eval_f1": 0.712166172106825,
      "eval_loss": 0.4479813575744629,
      "eval_precision": 0.72,
      "eval_recall": 0.7045009784735812,
      "eval_runtime": 28.0693,
      "eval_samples_per_second": 36.41,
      "eval_steps_per_second": 1.14,
      "step": 52
    },
    {
      "epoch": 2.0,
      "grad_norm": 8.902515411376953,
      "learning_rate": 8.981120615638859e-05,
      "loss": 0.4475,
      "step": 104
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7504892367906066,
      "eval_f1": 0.7578347578347578,
      "eval_loss": 0.42175495624542236,
      "eval_precision": 0.7361623616236163,
      "eval_recall": 0.7808219178082192,
      "eval_runtime": 27.9965,
      "eval_samples_per_second": 36.505,
      "eval_steps_per_second": 1.143,
      "step": 104
    },
    {
      "epoch": 3.0,
      "grad_norm": 4.328202724456787,
      "learning_rate": 4.4905603078194294e-05,
      "loss": 0.4233,
      "step": 156
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7201565557729941,
      "eval_f1": 0.6857142857142856,
      "eval_loss": 0.415389746427536,
      "eval_precision": 0.7819548872180451,
      "eval_recall": 0.6105675146771037,
      "eval_runtime": 28.1971,
      "eval_samples_per_second": 36.245,
      "eval_steps_per_second": 1.135,
      "step": 156
    }
  ],
  "logging_steps": 500,
  "max_steps": 208,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 60,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7020704368642945,
    "learning_rate": 0.00017962241231277718,
    "num_train_epochs": 4,
    "per_device_train_batch_size": 60,
    "temperature": 19
  }
}