run-28/checkpoint-642/trainer_state.json

{
  "best_metric": 0.7428236672524897,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-28/checkpoint-428",
  "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": 5.574616432189941,
      "learning_rate": 0.00036655879725586745,
      "loss": 0.5539,
      "step": 214
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7287639132981839,
      "eval_f1": 0.43742405832320774,
      "eval_loss": 0.5075973272323608,
      "eval_mcc": 0.3328904757496092,
      "eval_precision": 0.7086614173228346,
      "eval_recall": 0.3163444639718805,
      "eval_runtime": 3.1198,
      "eval_samples_per_second": 547.147,
      "eval_steps_per_second": 17.309,
      "step": 214
    },
    {
      "epoch": 2.0,
      "grad_norm": 2.6217339038848877,
      "learning_rate": 0.00031419325479074356,
      "loss": 0.498,
      "step": 428
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7428236672524897,
      "eval_f1": 0.5364308342133052,
      "eval_loss": 0.4864349961280823,
      "eval_mcc": 0.38309619293082,
      "eval_precision": 0.671957671957672,
      "eval_recall": 0.44639718804920914,
      "eval_runtime": 3.1068,
      "eval_samples_per_second": 549.439,
      "eval_steps_per_second": 17.381,
      "step": 428
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.2401552200317383,
      "learning_rate": 0.0002618277123256196,
      "loss": 0.4809,
      "step": 642
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7275922671353251,
      "eval_f1": 0.3638850889192886,
      "eval_loss": 0.49476319551467896,
      "eval_mcc": 0.33497438544889974,
      "eval_precision": 0.8209876543209876,
      "eval_recall": 0.23374340949033393,
      "eval_runtime": 3.1348,
      "eval_samples_per_second": 544.536,
      "eval_steps_per_second": 17.226,
      "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": 0.9057585906996103,
    "learning_rate": 0.0004189243397209914,
    "num_train_epochs": 8,
    "temperature": 28
  }
}