run-22/checkpoint-219/trainer_state.json

{
  "best_metric": 0.7534246575342466,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-22/checkpoint-219",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 219,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 0.7895054817199707,
      "learning_rate": 4.53929460427656e-05,
      "loss": 0.592,
      "step": 73
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.538160469667319,
      "eval_f1": 0.1945392491467577,
      "eval_loss": 0.5740381479263306,
      "eval_precision": 0.76,
      "eval_recall": 0.11154598825831702,
      "eval_runtime": 28.0661,
      "eval_samples_per_second": 36.414,
      "eval_steps_per_second": 1.14,
      "step": 73
    },
    {
      "epoch": 2.0,
      "grad_norm": 3.0389113426208496,
      "learning_rate": 2.26964730213828e-05,
      "loss": 0.5416,
      "step": 146
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7240704500978473,
      "eval_f1": 0.7019027484143764,
      "eval_loss": 0.4974684715270996,
      "eval_precision": 0.7632183908045977,
      "eval_recall": 0.649706457925636,
      "eval_runtime": 28.2061,
      "eval_samples_per_second": 36.233,
      "eval_steps_per_second": 1.135,
      "step": 146
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.835768699645996,
      "learning_rate": 0.0,
      "loss": 0.4934,
      "step": 219
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7534246575342466,
      "eval_f1": 0.7692307692307693,
      "eval_loss": 0.47661900520324707,
      "eval_precision": 0.7228915662650602,
      "eval_recall": 0.821917808219178,
      "eval_runtime": 28.9628,
      "eval_samples_per_second": 35.287,
      "eval_steps_per_second": 1.105,
      "step": 219
    }
  ],
  "logging_steps": 500,
  "max_steps": 219,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 42,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.8218865571627406,
    "learning_rate": 6.80894190641484e-05,
    "num_train_epochs": 3,
    "per_device_train_batch_size": 42,
    "temperature": 18
  }
}