run-13/checkpoint-144/trainer_state.json

{
  "best_metric": 0.8091976516634051,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-13/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": 1.1932880878448486,
      "learning_rate": 0.00031905315423317427,
      "loss": 0.5324,
      "step": 48
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7514677103718199,
      "eval_f1": 0.7897350993377483,
      "eval_loss": 0.4841874837875366,
      "eval_precision": 0.6843615494978479,
      "eval_recall": 0.9334637964774951,
      "eval_runtime": 27.3845,
      "eval_samples_per_second": 37.32,
      "eval_steps_per_second": 0.584,
      "step": 48
    },
    {
      "epoch": 2.0,
      "grad_norm": 5.759732723236084,
      "learning_rate": 0.0002552425233865394,
      "loss": 0.4624,
      "step": 96
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7720156555772995,
      "eval_f1": 0.762487257900102,
      "eval_loss": 0.43731480836868286,
      "eval_precision": 0.7957446808510639,
      "eval_recall": 0.7318982387475538,
      "eval_runtime": 27.3758,
      "eval_samples_per_second": 37.332,
      "eval_steps_per_second": 0.584,
      "step": 96
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.5165303945541382,
      "learning_rate": 0.00019143189253990455,
      "loss": 0.4346,
      "step": 144
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8091976516634051,
      "eval_f1": 0.8133971291866028,
      "eval_loss": 0.4113065004348755,
      "eval_precision": 0.795880149812734,
      "eval_recall": 0.8317025440313112,
      "eval_runtime": 26.8327,
      "eval_samples_per_second": 38.088,
      "eval_steps_per_second": 0.596,
      "step": 144
    }
  ],
  "logging_steps": 500,
  "max_steps": 288,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.804266474524187,
    "learning_rate": 0.0003828637850798091,
    "num_train_epochs": 6,
    "temperature": 12
  }
}