run-8/checkpoint-288/trainer_state.json

{
  "best_metric": 0.761252446183953,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-8/checkpoint-288",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 288,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.6551002860069275,
      "learning_rate": 6.801658733829082e-05,
      "loss": 0.4365,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.662426614481409,
      "eval_f1": 0.558258642765685,
      "eval_loss": 0.3849833011627197,
      "eval_precision": 0.8074074074074075,
      "eval_recall": 0.42661448140900193,
      "eval_runtime": 25.2796,
      "eval_samples_per_second": 40.428,
      "eval_steps_per_second": 1.266,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.109588384628296,
      "learning_rate": 5.9514513921004466e-05,
      "loss": 0.3828,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7045009784735812,
      "eval_f1": 0.6544622425629291,
      "eval_loss": 0.3611763119697571,
      "eval_precision": 0.7878787878787878,
      "eval_recall": 0.5596868884540117,
      "eval_runtime": 25.4122,
      "eval_samples_per_second": 40.217,
      "eval_steps_per_second": 1.259,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.4850609302520752,
      "learning_rate": 5.101244050371811e-05,
      "loss": 0.3691,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.761252446183953,
      "eval_f1": 0.7525354969574035,
      "eval_loss": 0.35745102167129517,
      "eval_precision": 0.7810526315789473,
      "eval_recall": 0.726027397260274,
      "eval_runtime": 25.6214,
      "eval_samples_per_second": 39.889,
      "eval_steps_per_second": 1.249,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 864,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 9,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.5720238366992816,
    "learning_rate": 7.651866075557717e-05,
    "num_train_epochs": 9,
    "temperature": 25
  }
}