run-41/checkpoint-288/trainer_state.json

{
  "best_metric": 0.824853228962818,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-41/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": 2.904797315597534,
      "learning_rate": 0.0001800620290766639,
      "loss": 0.5678,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7749510763209393,
      "eval_f1": 0.8060708263069141,
      "eval_loss": 0.47195959091186523,
      "eval_precision": 0.7081481481481482,
      "eval_recall": 0.9354207436399217,
      "eval_runtime": 32.7213,
      "eval_samples_per_second": 31.233,
      "eval_steps_per_second": 0.978,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 4.956325054168701,
      "learning_rate": 9.003101453833196e-05,
      "loss": 0.4548,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8199608610567515,
      "eval_f1": 0.8327272727272728,
      "eval_loss": 0.42627328634262085,
      "eval_precision": 0.7775891341256367,
      "eval_recall": 0.8962818003913894,
      "eval_runtime": 30.8838,
      "eval_samples_per_second": 33.092,
      "eval_steps_per_second": 1.036,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 4.930178165435791,
      "learning_rate": 0.0,
      "loss": 0.4292,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.824853228962818,
      "eval_f1": 0.8385933273219117,
      "eval_loss": 0.4179724156856537,
      "eval_precision": 0.7775919732441472,
      "eval_recall": 0.9099804305283757,
      "eval_runtime": 31.635,
      "eval_samples_per_second": 32.306,
      "eval_steps_per_second": 1.012,
      "step": 288
    }
  ],
  "logging_steps": 500,
  "max_steps": 288,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.9482965338753422,
    "learning_rate": 0.0002700930436149959,
    "num_train_epochs": 3,
    "temperature": 9
  }
}