run-0/checkpoint-288/trainer_state.json

{
  "best_metric": 0.6790606653620352,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-0/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": 0.4804088771343231,
      "learning_rate": 2.9818131608660274e-05,
      "loss": 0.4786,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5,
      "eval_f1": 0.0,
      "eval_loss": 0.4664871394634247,
      "eval_precision": 0.0,
      "eval_recall": 0.0,
      "eval_runtime": 28.9797,
      "eval_samples_per_second": 35.266,
      "eval_steps_per_second": 1.104,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.77193021774292,
      "learning_rate": 1.9878754405773516e-05,
      "loss": 0.4481,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.5958904109589042,
      "eval_f1": 0.37707390648567113,
      "eval_loss": 0.4186549484729767,
      "eval_precision": 0.8223684210526315,
      "eval_recall": 0.2446183953033268,
      "eval_runtime": 28.789,
      "eval_samples_per_second": 35.5,
      "eval_steps_per_second": 1.112,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.5943424701690674,
      "learning_rate": 9.939377202886758e-06,
      "loss": 0.4139,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.6790606653620352,
      "eval_f1": 0.6019417475728155,
      "eval_loss": 0.39717963337898254,
      "eval_precision": 0.792332268370607,
      "eval_recall": 0.48532289628180036,
      "eval_runtime": 29.1776,
      "eval_samples_per_second": 35.027,
      "eval_steps_per_second": 1.097,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 384,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.6186723691968887,
    "learning_rate": 3.975750881154703e-05,
    "num_train_epochs": 4,
    "temperature": 6
  }
}