run-33/checkpoint-288/trainer_state.json

{
  "best_metric": 0.8238747553816047,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-33/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": 2.87579345703125,
      "learning_rate": 0.00016441875793172356,
      "loss": 0.5654,
      "step": 96
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7896281800391389,
      "eval_f1": 0.816396242527754,
      "eval_loss": 0.4605298936367035,
      "eval_precision": 0.7242424242424242,
      "eval_recall": 0.9354207436399217,
      "eval_runtime": 29.158,
      "eval_samples_per_second": 35.05,
      "eval_steps_per_second": 1.097,
      "step": 96
    },
    {
      "epoch": 2.0,
      "grad_norm": 5.452133655548096,
      "learning_rate": 0.0001096125052878157,
      "loss": 0.4617,
      "step": 192
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8101761252446184,
      "eval_f1": 0.829225352112676,
      "eval_loss": 0.43088027834892273,
      "eval_precision": 0.7536,
      "eval_recall": 0.9217221135029354,
      "eval_runtime": 29.758,
      "eval_samples_per_second": 34.344,
      "eval_steps_per_second": 1.075,
      "step": 192
    },
    {
      "epoch": 3.0,
      "grad_norm": 4.057122230529785,
      "learning_rate": 5.480625264390785e-05,
      "loss": 0.4328,
      "step": 288
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8238747553816047,
      "eval_f1": 0.832089552238806,
      "eval_loss": 0.4205124080181122,
      "eval_precision": 0.7950089126559715,
      "eval_recall": 0.87279843444227,
      "eval_runtime": 29.1506,
      "eval_samples_per_second": 35.059,
      "eval_steps_per_second": 1.098,
      "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": 0.9211741662812639,
    "learning_rate": 0.0002192250105756314,
    "num_train_epochs": 4,
    "temperature": 2
  }
}