run-1/checkpoint-288/trainer_state.json

{
  "best_metric": 0.8013698630136986,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-1/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": 1.0269759893417358,
      "learning_rate": 5.9043684647359026e-05,
      "loss": 0.5339,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7035225048923679,
      "eval_f1": 0.6651933701657459,
      "eval_loss": 0.4725688099861145,
      "eval_precision": 0.7639593908629442,
      "eval_recall": 0.589041095890411,
      "eval_runtime": 28.4268,
      "eval_samples_per_second": 35.952,
      "eval_steps_per_second": 1.126,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.4684391021728516,
      "learning_rate": 4.920307053946586e-05,
      "loss": 0.4589,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7563600782778865,
      "eval_f1": 0.7537091988130563,
      "eval_loss": 0.4252457022666931,
      "eval_precision": 0.762,
      "eval_recall": 0.7455968688845401,
      "eval_runtime": 28.3965,
      "eval_samples_per_second": 35.99,
      "eval_steps_per_second": 1.127,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.83766770362854,
      "learning_rate": 3.9362456431572686e-05,
      "loss": 0.4336,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8013698630136986,
      "eval_f1": 0.8122109158186863,
      "eval_loss": 0.41349494457244873,
      "eval_precision": 0.7701754385964912,
      "eval_recall": 0.8590998043052838,
      "eval_runtime": 28.5368,
      "eval_samples_per_second": 35.813,
      "eval_steps_per_second": 1.121,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 672,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 7,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7341280696937652,
    "learning_rate": 6.88842987552522e-05,
    "num_train_epochs": 7,
    "temperature": 8
  }
}