run-3/checkpoint-144/trainer_state.json

{
  "best_metric": 0.6526418786692759,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-3/checkpoint-96",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 144,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.7208284735679626,
      "learning_rate": 0.00012206653312994039,
      "loss": 0.4111,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5792563600782779,
      "eval_f1": 0.3343653250773994,
      "eval_loss": 0.375780314207077,
      "eval_precision": 0.8,
      "eval_recall": 0.21135029354207435,
      "eval_runtime": 28.363,
      "eval_samples_per_second": 36.033,
      "eval_steps_per_second": 0.564,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.6850707530975342,
      "learning_rate": 8.137768875329359e-05,
      "loss": 0.3675,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6526418786692759,
      "eval_f1": 0.5285524568393094,
      "eval_loss": 0.3478649854660034,
      "eval_precision": 0.8223140495867769,
      "eval_recall": 0.38943248532289626,
      "eval_runtime": 27.159,
      "eval_samples_per_second": 37.63,
      "eval_steps_per_second": 0.589,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.8926218748092651,
      "learning_rate": 4.0688844376646796e-05,
      "loss": 0.3521,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.6506849315068494,
      "eval_f1": 0.5333333333333332,
      "eval_loss": 0.3429274559020996,
      "eval_precision": 0.8031496062992126,
      "eval_recall": 0.39921722113502933,
      "eval_runtime": 27.1326,
      "eval_samples_per_second": 37.667,
      "eval_steps_per_second": 0.59,
      "step": 144
    }
  ],
  "logging_steps": 500,
  "max_steps": 192,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.5259618440624823,
    "learning_rate": 0.00016275537750658718,
    "num_train_epochs": 4,
    "temperature": 27
  }
}