run-18/checkpoint-642/trainer_state.json

{
  "best_metric": 0.6666666666666666,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-18/checkpoint-214",
  "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": 2.14192795753479,
      "learning_rate": 3.504846464658413e-05,
      "loss": 0.2876,
      "step": 214
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6666666666666666,
      "eval_f1": 0.0,
      "eval_loss": 0.2669273614883423,
      "eval_mcc": 0.0,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 3.1234,
      "eval_samples_per_second": 546.513,
      "eval_steps_per_second": 17.289,
      "step": 214
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.7823044061660767,
      "learning_rate": 3.0667406565761116e-05,
      "loss": 0.2659,
      "step": 428
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6666666666666666,
      "eval_f1": 0.0,
      "eval_loss": 0.25948211550712585,
      "eval_mcc": 0.0,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 3.151,
      "eval_samples_per_second": 541.737,
      "eval_steps_per_second": 17.138,
      "step": 428
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.1866549253463745,
      "learning_rate": 2.6286348484938096e-05,
      "loss": 0.2602,
      "step": 642
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.6666666666666666,
      "eval_f1": 0.0,
      "eval_loss": 0.2597707509994507,
      "eval_mcc": 0.0,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 3.1289,
      "eval_samples_per_second": 545.561,
      "eval_steps_per_second": 17.259,
      "step": 642
    }
  ],
  "logging_steps": 500,
  "max_steps": 1926,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 9,
  "save_steps": 500,
  "total_flos": 1574326993320.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.4076783392967324,
    "learning_rate": 3.942952272740715e-05,
    "num_train_epochs": 9,
    "temperature": 30
  }
}