{
  "best_metric": 0.8038990825688074,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-2/checkpoint-2108",
  "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": 6.471368312835693,
      "learning_rate": 2.8549539646703335e-05,
      "loss": 2.1389,
      "step": 527
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7763761467889908,
      "eval_loss": 1.5801109075546265,
      "eval_runtime": 2.8093,
      "eval_samples_per_second": 310.393,
      "eval_steps_per_second": 2.492,
      "step": 527
    },
    {
      "epoch": 2.0,
      "grad_norm": Infinity,
      "learning_rate": 1.906914223081513e-05,
      "loss": 1.3352,
      "step": 1054
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7993119266055045,
      "eval_loss": 1.3590723276138306,
      "eval_runtime": 2.8218,
      "eval_samples_per_second": 309.021,
      "eval_steps_per_second": 2.481,
      "step": 1054
    },
    {
      "epoch": 3.0,
      "grad_norm": 37.449886322021484,
      "learning_rate": 9.552629015247352e-06,
      "loss": 1.0924,
      "step": 1581
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7981651376146789,
      "eval_loss": 1.310165286064148,
      "eval_runtime": 2.8142,
      "eval_samples_per_second": 309.857,
      "eval_steps_per_second": 2.487,
      "step": 1581
    },
    {
      "epoch": 4.0,
      "grad_norm": 20.399564743041992,
      "learning_rate": 3.611579967957411e-08,
      "loss": 0.9927,
      "step": 2108
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8038990825688074,
      "eval_loss": 1.2949246168136597,
      "eval_runtime": 2.8058,
      "eval_samples_per_second": 310.787,
      "eval_steps_per_second": 2.495,
      "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": 0.5278139834631775,
    "learning_rate": 3.806605286227111e-05,
    "num_train_epochs": 4,
    "temperature": 30
  }
}