{
  "best_metric": 0.8715722120658135,
  "best_model_checkpoint": "distilbert-hate_speech18\\run-5\\checkpoint-1916",
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 1916,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 1.4395797527808295e-06,
      "loss": 0.2516,
      "step": 479
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8624314442413162,
      "eval_loss": 0.24848461151123047,
      "eval_runtime": 9.4673,
      "eval_samples_per_second": 231.112,
      "eval_steps_per_second": 7.288,
      "step": 479
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.1516638022246633e-06,
      "loss": 0.2482,
      "step": 958
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8638025594149908,
      "eval_loss": 0.2472020387649536,
      "eval_runtime": 9.4556,
      "eval_samples_per_second": 231.396,
      "eval_steps_per_second": 7.297,
      "step": 958
    },
    {
      "epoch": 3.0,
      "learning_rate": 8.637478516684976e-07,
      "loss": 0.2469,
      "step": 1437
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8688299817184644,
      "eval_loss": 0.24652723968029022,
      "eval_runtime": 9.4688,
      "eval_samples_per_second": 231.075,
      "eval_steps_per_second": 7.287,
      "step": 1437
    },
    {
      "epoch": 4.0,
      "learning_rate": 5.758319011123317e-07,
      "loss": 0.2461,
      "step": 1916
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8715722120658135,
      "eval_loss": 0.24602682888507843,
      "eval_runtime": 9.478,
      "eval_samples_per_second": 230.849,
      "eval_steps_per_second": 7.28,
      "step": 1916
    }
  ],
  "logging_steps": 500,
  "max_steps": 2874,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 500,
  "total_flos": 514695596916480.0,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.19436109341303698,
    "learning_rate": 1.7274957033369952e-06,
    "num_train_epochs": 6,
    "per_device_eval_batch_size": 32,
    "per_device_train_batch_size": 16,
    "temperature": 12
  }
}