run-6/checkpoint-144/trainer_state.json

{
  "best_metric": 0.8287671232876712,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-6/checkpoint-144",
  "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": 1.4916555881500244,
      "learning_rate": 0.00048113016952618413,
      "loss": 0.5799,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7338551859099804,
      "eval_f1": 0.7854889589905363,
      "eval_loss": 0.5085063576698303,
      "eval_precision": 0.6578599735799208,
      "eval_recall": 0.974559686888454,
      "eval_runtime": 27.0072,
      "eval_samples_per_second": 37.842,
      "eval_steps_per_second": 0.592,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 8.242779731750488,
      "learning_rate": 0.00024056508476309206,
      "loss": 0.4762,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.812133072407045,
      "eval_f1": 0.8254545454545454,
      "eval_loss": 0.43319791555404663,
      "eval_precision": 0.7707979626485568,
      "eval_recall": 0.8884540117416829,
      "eval_runtime": 27.355,
      "eval_samples_per_second": 37.361,
      "eval_steps_per_second": 0.585,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.5861244201660156,
      "learning_rate": 0.0,
      "loss": 0.4211,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8287671232876712,
      "eval_f1": 0.8444444444444446,
      "eval_loss": 0.40428370237350464,
      "eval_precision": 0.7736156351791531,
      "eval_recall": 0.9295499021526419,
      "eval_runtime": 27.6148,
      "eval_samples_per_second": 37.009,
      "eval_steps_per_second": 0.579,
      "step": 144
    }
  ],
  "logging_steps": 500,
  "max_steps": 144,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.9704925152657605,
    "learning_rate": 0.0007216952542892762,
    "num_train_epochs": 3,
    "temperature": 28
  }
}