{
  "best_metric": 0.8314220183486238,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-41/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": 13.133516311645508,
      "learning_rate": 0.00019951632086846584,
      "loss": 0.9141,
      "step": 527
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8314220183486238,
      "eval_loss": 0.9435493350028992,
      "eval_runtime": 2.8138,
      "eval_samples_per_second": 309.897,
      "eval_steps_per_second": 2.488,
      "step": 527
    },
    {
      "epoch": 2.0,
      "grad_norm": 6.979046821594238,
      "learning_rate": 0.00013301088057897722,
      "loss": 0.4473,
      "step": 1054
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8268348623853211,
      "eval_loss": 1.0495487451553345,
      "eval_runtime": 2.8143,
      "eval_samples_per_second": 309.848,
      "eval_steps_per_second": 2.487,
      "step": 1054
    },
    {
      "epoch": 3.0,
      "grad_norm": 42.47988510131836,
      "learning_rate": 6.650544028948861e-05,
      "loss": 0.3269,
      "step": 1581
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8176605504587156,
      "eval_loss": 1.1648478507995605,
      "eval_runtime": 2.825,
      "eval_samples_per_second": 308.672,
      "eval_steps_per_second": 2.478,
      "step": 1581
    },
    {
      "epoch": 4.0,
      "grad_norm": 6.063735008239746,
      "learning_rate": 0.0,
      "loss": 0.2671,
      "step": 2108
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8222477064220184,
      "eval_loss": 1.1299054622650146,
      "eval_runtime": 2.8134,
      "eval_samples_per_second": 309.941,
      "eval_steps_per_second": 2.488,
      "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": 0.6708040997397258,
    "learning_rate": 0.00026602176115795445,
    "num_train_epochs": 4,
    "temperature": 9
  }
}