run-3/checkpoint-354/trainer_state.json

{
  "best_metric": 0.8082191780821918,
  "best_model_checkpoint": "tiny-bert-sst2-distilled/run-3/checkpoint-354",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 354,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 1.14125657081604,
      "learning_rate": 7.61866849595866e-05,
      "loss": 0.532,
      "step": 118
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7475538160469667,
      "eval_f1": 0.7575187969924811,
      "eval_loss": 0.4580899477005005,
      "eval_precision": 0.72875226039783,
      "eval_recall": 0.7886497064579256,
      "eval_runtime": 30.2456,
      "eval_samples_per_second": 33.79,
      "eval_steps_per_second": 1.058,
      "step": 118
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.8153949975967407,
      "learning_rate": 3.80933424797933e-05,
      "loss": 0.4543,
      "step": 236
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7984344422700587,
      "eval_f1": 0.8096118299445471,
      "eval_loss": 0.4280538260936737,
      "eval_precision": 0.7670753064798599,
      "eval_recall": 0.8571428571428571,
      "eval_runtime": 29.8013,
      "eval_samples_per_second": 34.294,
      "eval_steps_per_second": 1.074,
      "step": 236
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.0012309551239014,
      "learning_rate": 0.0,
      "loss": 0.4357,
      "step": 354
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8082191780821918,
      "eval_f1": 0.8246869409660107,
      "eval_loss": 0.4239581227302551,
      "eval_precision": 0.7594728171334432,
      "eval_recall": 0.9021526418786693,
      "eval_runtime": 30.1311,
      "eval_samples_per_second": 33.918,
      "eval_steps_per_second": 1.062,
      "step": 354
    }
  ],
  "logging_steps": 500,
  "max_steps": 354,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 707085591840.0,
  "train_batch_size": 26,
  "trial_name": null,
  "trial_params": {
    "alpha": 0.7708734866574575,
    "learning_rate": 0.00011428002743937991,
    "num_train_epochs": 3,
    "per_device_train_batch_size": 26,
    "temperature": 7
  }
}