{
  "best_metric": 0.8291284403669725,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-44/checkpoint-527",
  "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": 15.370396614074707,
      "learning_rate": 0.000199328116615274,
      "loss": 0.9859,
      "step": 527
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8291284403669725,
      "eval_loss": 1.0127737522125244,
      "eval_runtime": 2.815,
      "eval_samples_per_second": 309.77,
      "eval_steps_per_second": 2.487,
      "step": 527
    },
    {
      "epoch": 2.0,
      "grad_norm": 7.43062162399292,
      "learning_rate": 0.00013288541107684933,
      "loss": 0.4754,
      "step": 1054
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8268348623853211,
      "eval_loss": 1.1078842878341675,
      "eval_runtime": 2.8191,
      "eval_samples_per_second": 309.317,
      "eval_steps_per_second": 2.483,
      "step": 1054
    },
    {
      "epoch": 3.0,
      "grad_norm": 44.074180603027344,
      "learning_rate": 6.644270553842467e-05,
      "loss": 0.3475,
      "step": 1581
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8176605504587156,
      "eval_loss": 1.210789442062378,
      "eval_runtime": 2.8179,
      "eval_samples_per_second": 309.454,
      "eval_steps_per_second": 2.484,
      "step": 1581
    },
    {
      "epoch": 4.0,
      "grad_norm": 6.042885780334473,
      "learning_rate": 0.0,
      "loss": 0.2836,
      "step": 2108
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.823394495412844,
      "eval_loss": 1.17649245262146,
      "eval_runtime": 2.8084,
      "eval_samples_per_second": 310.498,
      "eval_steps_per_second": 2.493,
      "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": 0.6434351873812945,
    "learning_rate": 0.00026577082215369866,
    "num_train_epochs": 4,
    "temperature": 15
  }
}