run-12/checkpoint-642/trainer_state.json

{
  "best_metric": 0.6678383128295254,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-12/checkpoint-642",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 642,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.3038450479507446,
      "learning_rate": 8.466711582234029e-05,
      "loss": 0.1784,
      "step": 214
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6666666666666666,
      "eval_f1": 0.0,
      "eval_loss": 0.16210851073265076,
      "eval_mcc": 0.0,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 3.1313,
      "eval_samples_per_second": 545.14,
      "eval_steps_per_second": 17.245,
      "step": 214
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.5095393657684326,
      "learning_rate": 7.257181356200596e-05,
      "loss": 0.1644,
      "step": 428
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6660808435852372,
      "eval_f1": 0.003496503496503496,
      "eval_loss": 0.16058450937271118,
      "eval_mcc": 0.0,
      "eval_precision": 0.3333333333333333,
      "eval_recall": 0.0017574692442882249,
      "eval_runtime": 3.1478,
      "eval_samples_per_second": 542.278,
      "eval_steps_per_second": 17.155,
      "step": 428
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.4557552635669708,
      "learning_rate": 6.047651130167164e-05,
      "loss": 0.1626,
      "step": 642
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.6678383128295254,
      "eval_f1": 0.013913043478260868,
      "eval_loss": 0.16264592111110687,
      "eval_mcc": 0.0419960525565808,
      "eval_precision": 0.6666666666666666,
      "eval_recall": 0.007029876977152899,
      "eval_runtime": 3.1385,
      "eval_samples_per_second": 543.883,
      "eval_steps_per_second": 17.205,
      "step": 642
    }
  ],
  "logging_steps": 500,
  "max_steps": 1712,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 8,
  "save_steps": 500,
  "total_flos": 1574326993320.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.24525395176272902,
    "learning_rate": 9.676241808267462e-05,
    "num_train_epochs": 8,
    "temperature": 20
  }
}