run-4/checkpoint-670/trainer_state.json

{
  "best_metric": 0.5282404248888111,
  "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-4/checkpoint-670",
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 670,
  "is_hyper_param_search": true,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_loss": 0.4793247580528259,
      "eval_matthews_correlation": 0.41721186495955986,
      "eval_runtime": 0.7047,
      "eval_samples_per_second": 1480.003,
      "eval_steps_per_second": 93.653,
      "step": 134
    },
    {
      "epoch": 2.0,
      "eval_loss": 0.5243192315101624,
      "eval_matthews_correlation": 0.4857615494749571,
      "eval_runtime": 0.7767,
      "eval_samples_per_second": 1342.919,
      "eval_steps_per_second": 84.979,
      "step": 268
    },
    {
      "epoch": 3.0,
      "eval_loss": 0.6017581224441528,
      "eval_matthews_correlation": 0.5127204235005072,
      "eval_runtime": 0.7858,
      "eval_samples_per_second": 1327.328,
      "eval_steps_per_second": 83.992,
      "step": 402
    },
    {
      "epoch": 3.73,
      "grad_norm": 3.175116539001465,
      "learning_rate": 2.121179510811267e-05,
      "loss": 0.2876,
      "step": 500
    },
    {
      "epoch": 4.0,
      "eval_loss": 0.7295573949813843,
      "eval_matthews_correlation": 0.516230230689141,
      "eval_runtime": 0.7821,
      "eval_samples_per_second": 1333.555,
      "eval_steps_per_second": 84.386,
      "step": 536
    },
    {
      "epoch": 5.0,
      "eval_loss": 0.8349049687385559,
      "eval_matthews_correlation": 0.5282404248888111,
      "eval_runtime": 1.5,
      "eval_samples_per_second": 695.316,
      "eval_steps_per_second": 43.999,
      "step": 670
    }
  ],
  "logging_steps": 500,
  "max_steps": 670,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 210826263292620.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": {
    "learning_rate": 8.359942777903227e-05,
    "num_train_epochs": 5,
    "per_device_train_batch_size": 64,
    "seed": 24
  }
}