run-5/checkpoint-462/trainer_state.json

{
  "best_metric": 0.538160469667319,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-5/checkpoint-462",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 462,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.6735444664955139,
      "learning_rate": 1.0944822590536588e-05,
      "loss": 0.1064,
      "step": 154
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5,
      "eval_f1": 0.0,
      "eval_loss": 0.04401332512497902,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 28.8317,
      "eval_samples_per_second": 35.447,
      "eval_steps_per_second": 1.11,
      "step": 154
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.7938733696937561,
      "learning_rate": 9.728731191588077e-06,
      "loss": 0.0382,
      "step": 308
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.5097847358121331,
      "eval_f1": 0.05649717514124293,
      "eval_loss": 0.02585103176534176,
      "eval_precision": 0.75,
      "eval_recall": 0.029354207436399216,
      "eval_runtime": 29.1621,
      "eval_samples_per_second": 35.046,
      "eval_steps_per_second": 1.097,
      "step": 308
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.4618286192417145,
      "learning_rate": 8.512639792639567e-06,
      "loss": 0.027,
      "step": 462
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.538160469667319,
      "eval_f1": 0.1917808219178082,
      "eval_loss": 0.020071249455213547,
      "eval_precision": 0.7671232876712328,
      "eval_recall": 0.1095890410958904,
      "eval_runtime": 27.7471,
      "eval_samples_per_second": 36.833,
      "eval_steps_per_second": 1.153,
      "step": 462
    }
  ],
  "logging_steps": 500,
  "max_steps": 1540,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 20,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.007207939952405096,
    "learning_rate": 1.2160913989485096e-05,
    "num_train_epochs": 10,
    "per_device_train_batch_size": 20,
    "temperature": 25
  }
}