run-21/checkpoint-144/trainer_state.json

{
  "best_metric": 0.8277886497064579,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-21/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": 1.955494999885559,
      "learning_rate": 0.00045025211447661823,
      "loss": 0.583,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7544031311154599,
      "eval_f1": 0.7957689178193653,
      "eval_loss": 0.48823896050453186,
      "eval_precision": 0.6810584958217271,
      "eval_recall": 0.9569471624266145,
      "eval_runtime": 27.5121,
      "eval_samples_per_second": 37.147,
      "eval_steps_per_second": 0.582,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 5.347569942474365,
      "learning_rate": 0.0003001680763177455,
      "loss": 0.4846,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.815068493150685,
      "eval_f1": 0.8205128205128206,
      "eval_loss": 0.4311129152774811,
      "eval_precision": 0.7970479704797048,
      "eval_recall": 0.8454011741682974,
      "eval_runtime": 27.1948,
      "eval_samples_per_second": 37.581,
      "eval_steps_per_second": 0.588,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.818072557449341,
      "learning_rate": 0.00015008403815887274,
      "loss": 0.4238,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8277886497064579,
      "eval_f1": 0.8361266294227189,
      "eval_loss": 0.41256558895111084,
      "eval_precision": 0.7975133214920072,
      "eval_recall": 0.8786692759295499,
      "eval_runtime": 27.3265,
      "eval_samples_per_second": 37.4,
      "eval_steps_per_second": 0.586,
      "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": 0.9858291446272915,
    "learning_rate": 0.000600336152635491,
    "num_train_epochs": 4,
    "temperature": 30
  }
}