{
  "best_metric": 0.8302752293577982,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-5/checkpoint-1054",
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 2108,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 8.730439186096191,
      "learning_rate": 8.635300806475536e-05,
      "loss": 1.4959,
      "step": 527
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8061926605504587,
      "eval_loss": 1.2476624250411987,
      "eval_runtime": 2.8128,
      "eval_samples_per_second": 310.007,
      "eval_steps_per_second": 2.489,
      "step": 527
    },
    {
      "epoch": 2.0,
      "grad_norm": 17.52631950378418,
      "learning_rate": 6.476475604856652e-05,
      "loss": 0.765,
      "step": 1054
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8302752293577982,
      "eval_loss": 1.193213939666748,
      "eval_runtime": 2.8129,
      "eval_samples_per_second": 310.006,
      "eval_steps_per_second": 2.489,
      "step": 1054
    },
    {
      "epoch": 3.0,
      "grad_norm": NaN,
      "learning_rate": 4.3217468457455834e-05,
      "loss": 0.591,
      "step": 1581
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8245412844036697,
      "eval_loss": 1.2127447128295898,
      "eval_runtime": 2.8189,
      "eval_samples_per_second": 309.344,
      "eval_steps_per_second": 2.483,
      "step": 1581
    },
    {
      "epoch": 4.0,
      "grad_norm": 9.946269035339355,
      "learning_rate": 2.1629216441266998e-05,
      "loss": 0.5048,
      "step": 2108
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8302752293577982,
      "eval_loss": 1.2008661031723022,
      "eval_runtime": 2.8198,
      "eval_samples_per_second": 309.246,
      "eval_steps_per_second": 2.482,
      "step": 2108
    }
  ],
  "logging_steps": 500,
  "max_steps": 2635,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 32337743850420.0,
  "train_batch_size": 128,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.553843836109961,
    "learning_rate": 0.0001079412600809442,
    "num_train_epochs": 5,
    "temperature": 29
  }
}