run-10/checkpoint-642/trainer_state.json

{
  "best_metric": 0.6690099589923842,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-10/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": 1.0974807739257812,
      "learning_rate": 0.0004326089412215258,
      "loss": 0.1568,
      "step": 214
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6654950205038078,
      "eval_f1": 0.02393162393162393,
      "eval_loss": 0.14753811061382294,
      "eval_mcc": 0.021494295311278402,
      "eval_precision": 0.4375,
      "eval_recall": 0.012302284710017574,
      "eval_runtime": 3.1372,
      "eval_samples_per_second": 544.118,
      "eval_steps_per_second": 17.213,
      "step": 214
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.3948715031147003,
      "learning_rate": 0.00038454128108580076,
      "loss": 0.1489,
      "step": 428
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6690099589923842,
      "eval_f1": 0.053601340033500824,
      "eval_loss": 0.14619438350200653,
      "eval_mcc": 0.06522448130722841,
      "eval_precision": 0.5714285714285714,
      "eval_recall": 0.028119507908611598,
      "eval_runtime": 3.1687,
      "eval_samples_per_second": 538.707,
      "eval_steps_per_second": 17.042,
      "step": 428
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.6567928791046143,
      "learning_rate": 0.00033647362095007565,
      "loss": 0.1472,
      "step": 642
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.667252489748096,
      "eval_f1": 0.030716723549488054,
      "eval_loss": 0.1447131335735321,
      "eval_mcc": 0.04171739573435696,
      "eval_precision": 0.5294117647058824,
      "eval_recall": 0.015817223198594025,
      "eval_runtime": 3.1648,
      "eval_samples_per_second": 539.364,
      "eval_steps_per_second": 17.062,
      "step": 642
    }
  ],
  "logging_steps": 500,
  "max_steps": 2140,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 1574326993320.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.22222976837593644,
    "learning_rate": 0.0004806766013572509,
    "num_train_epochs": 10,
    "temperature": 30
  }
}