run-2/checkpoint-288/trainer_state.json

{
  "best_metric": 0.5391389432485323,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-2/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": 0.3969263434410095,
      "learning_rate": 3.130245712253287e-05,
      "loss": 0.081,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.512720156555773,
      "eval_f1": 0.08118081180811808,
      "eval_loss": 0.03750595450401306,
      "eval_precision": 0.7096774193548387,
      "eval_recall": 0.043052837573385516,
      "eval_runtime": 25.4915,
      "eval_samples_per_second": 40.092,
      "eval_steps_per_second": 1.255,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.2545386254787445,
      "learning_rate": 1.5651228561266437e-05,
      "loss": 0.0372,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.538160469667319,
      "eval_f1": 0.1945392491467577,
      "eval_loss": 0.030006052926182747,
      "eval_precision": 0.76,
      "eval_recall": 0.11154598825831702,
      "eval_runtime": 26.0378,
      "eval_samples_per_second": 39.251,
      "eval_steps_per_second": 1.229,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.24247972667217255,
      "learning_rate": 0.0,
      "loss": 0.0336,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.5391389432485323,
      "eval_f1": 0.19487179487179485,
      "eval_loss": 0.028951693326234818,
      "eval_precision": 0.7702702702702703,
      "eval_recall": 0.11154598825831702,
      "eval_runtime": 26.0257,
      "eval_samples_per_second": 39.269,
      "eval_steps_per_second": 1.23,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 288,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.022604047006827055,
    "learning_rate": 4.6953685683799313e-05,
    "num_train_epochs": 3,
    "temperature": 5
  }
}