{
  "best_metric": 0.6653465346534654,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-1/checkpoint-1188",
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 1188,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.3426679372787476,
      "learning_rate": 0.0003560526201910554,
      "loss": 0.6371,
      "step": 297
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5841584158415841,
      "eval_loss": 0.637409508228302,
      "eval_runtime": 56.7978,
      "eval_samples_per_second": 8.891,
      "eval_steps_per_second": 0.282,
      "step": 297
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.018396258354187,
      "learning_rate": 0.0002373684134607036,
      "loss": 0.603,
      "step": 594
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6277227722772277,
      "eval_loss": 0.6017976403236389,
      "eval_runtime": 55.6441,
      "eval_samples_per_second": 9.076,
      "eval_steps_per_second": 0.288,
      "step": 594
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.4982593059539795,
      "learning_rate": 0.0001186842067303518,
      "loss": 0.588,
      "step": 891
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.6495049504950495,
      "eval_loss": 0.5974766612052917,
      "eval_runtime": 55.1255,
      "eval_samples_per_second": 9.161,
      "eval_steps_per_second": 0.29,
      "step": 891
    },
    {
      "epoch": 4.0,
      "grad_norm": 2.6116485595703125,
      "learning_rate": 0.0,
      "loss": 0.5769,
      "step": 1188
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.6653465346534654,
      "eval_loss": 0.5864288806915283,
      "eval_runtime": 56.8255,
      "eval_samples_per_second": 8.887,
      "eval_steps_per_second": 0.282,
      "step": 1188
    }
  ],
  "logging_steps": 500,
  "max_steps": 1188,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 3111373132560.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.900582717199523,
    "learning_rate": 0.0004747368269214072,
    "num_train_epochs": 4,
    "temperature": 18
  }
}