{
  "best_metric": 0.8360091743119266,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-18/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": 9.684267044067383,
      "learning_rate": 0.00022228195193198518,
      "loss": 0.8996,
      "step": 527
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8337155963302753,
      "eval_loss": 0.9537320733070374,
      "eval_runtime": 2.8076,
      "eval_samples_per_second": 310.59,
      "eval_steps_per_second": 2.493,
      "step": 527
    },
    {
      "epoch": 2.0,
      "grad_norm": 6.183145523071289,
      "learning_rate": 0.00014818796795465679,
      "loss": 0.4363,
      "step": 1054
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8360091743119266,
      "eval_loss": 1.0877610445022583,
      "eval_runtime": 2.8149,
      "eval_samples_per_second": 309.776,
      "eval_steps_per_second": 2.487,
      "step": 1054
    },
    {
      "epoch": 3.0,
      "grad_norm": 37.144012451171875,
      "learning_rate": 7.409398397732839e-05,
      "loss": 0.3144,
      "step": 1581
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.823394495412844,
      "eval_loss": 1.1875492334365845,
      "eval_runtime": 2.8238,
      "eval_samples_per_second": 308.803,
      "eval_steps_per_second": 2.479,
      "step": 1581
    },
    {
      "epoch": 4.0,
      "grad_norm": 4.61117696762085,
      "learning_rate": 0.0,
      "loss": 0.2534,
      "step": 2108
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8279816513761468,
      "eval_loss": 1.1564440727233887,
      "eval_runtime": 2.8072,
      "eval_samples_per_second": 310.627,
      "eval_steps_per_second": 2.494,
      "step": 2108
    }
  ],
  "logging_steps": 500,
  "max_steps": 2108,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 32337743850420.0,
  "train_batch_size": 128,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.6751548126223466,
    "learning_rate": 0.00029637593590931357,
    "num_train_epochs": 4,
    "temperature": 11
  }
}