run-31/checkpoint-144/trainer_state.json

{
  "best_metric": 0.8268101761252447,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-31/checkpoint-144",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 144,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.7282975912094116,
      "learning_rate": 0.00030014515517910593,
      "loss": 0.5971,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7446183953033269,
      "eval_f1": 0.7900241351568785,
      "eval_loss": 0.5055608749389648,
      "eval_precision": 0.6707650273224044,
      "eval_recall": 0.9608610567514677,
      "eval_runtime": 27.0653,
      "eval_samples_per_second": 37.761,
      "eval_steps_per_second": 0.591,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 9.983665466308594,
      "learning_rate": 0.00020009677011940396,
      "loss": 0.4874,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8072407045009785,
      "eval_f1": 0.8089233753637245,
      "eval_loss": 0.45225080847740173,
      "eval_precision": 0.801923076923077,
      "eval_recall": 0.8160469667318982,
      "eval_runtime": 27.1905,
      "eval_samples_per_second": 37.587,
      "eval_steps_per_second": 0.588,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.998225450515747,
      "learning_rate": 0.00010004838505970198,
      "loss": 0.4428,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8268101761252447,
      "eval_f1": 0.8353488372093023,
      "eval_loss": 0.4180435538291931,
      "eval_precision": 0.7960992907801419,
      "eval_recall": 0.8786692759295499,
      "eval_runtime": 26.8998,
      "eval_samples_per_second": 37.993,
      "eval_steps_per_second": 0.595,
      "step": 144
    }
  ],
  "logging_steps": 500,
  "max_steps": 192,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.9938600685844597,
    "learning_rate": 0.0004001935402388079,
    "num_train_epochs": 4,
    "temperature": 28
  }
}