training_checkpoints/checkpoint-803/trainer_state.json

{
  "best_metric": 0.4824234674423381,
  "best_model_checkpoint": "legal-bert-lora/checkpoint-803",
  "epoch": 4.995334370139969,
  "eval_steps": 500,
  "global_step": 803,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.6374903175832688,
      "eval_f1_macro": 0.3394539856434376,
      "eval_f1_micro": 0.6374903175832688,
      "eval_f1_weighted": 0.5765523433851403,
      "eval_loss": 1.3127473592758179,
      "eval_macro_fpr": 0.040137717078130394,
      "eval_macro_sensitivity": 0.4017115530536667,
      "eval_macro_specificity": 0.9725650393599338,
      "eval_precision": 0.5466227876047401,
      "eval_precision_macro": 0.31141306924471374,
      "eval_recall": 0.6374903175832688,
      "eval_recall_macro": 0.4017115530536667,
      "eval_runtime": 44.4128,
      "eval_samples_per_second": 29.068,
      "eval_steps_per_second": 3.648,
      "eval_weighted_fpr": 0.0390325271059216,
      "eval_weighted_sensitivity": 0.6374903175832688,
      "eval_weighted_specificity": 0.950985272815738,
      "step": 160
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7273431448489543,
      "eval_f1_macro": 0.3969800602789428,
      "eval_f1_micro": 0.7273431448489543,
      "eval_f1_weighted": 0.6960627383550935,
      "eval_loss": 0.8160030841827393,
      "eval_macro_fpr": 0.026417996388603746,
      "eval_macro_sensitivity": 0.450135625445356,
      "eval_macro_specificity": 0.9796396430402219,
      "eval_precision": 0.6997929823446861,
      "eval_precision_macro": 0.39453164470926966,
      "eval_recall": 0.7273431448489543,
      "eval_recall_macro": 0.450135625445356,
      "eval_runtime": 34.3148,
      "eval_samples_per_second": 37.622,
      "eval_steps_per_second": 4.721,
      "eval_weighted_fpr": 0.026077937472218107,
      "eval_weighted_sensitivity": 0.7273431448489543,
      "eval_weighted_specificity": 0.9672515007543734,
      "step": 321
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7443841982958946,
      "eval_f1_macro": 0.4210575074216338,
      "eval_f1_micro": 0.7443841982958946,
      "eval_f1_weighted": 0.7223603623418946,
      "eval_loss": 0.7322217226028442,
      "eval_macro_fpr": 0.024451015698283406,
      "eval_macro_sensitivity": 0.46251578703813245,
      "eval_macro_specificity": 0.9809603839363172,
      "eval_precision": 0.711002444877192,
      "eval_precision_macro": 0.4030436676815596,
      "eval_recall": 0.7443841982958946,
      "eval_recall_macro": 0.46251578703813245,
      "eval_runtime": 32.8708,
      "eval_samples_per_second": 39.275,
      "eval_steps_per_second": 4.928,
      "eval_weighted_fpr": 0.023940800928612886,
      "eval_weighted_sensitivity": 0.7443841982958946,
      "eval_weighted_specificity": 0.9700215607488639,
      "step": 482
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.958333333333333e-05,
      "loss": 1.2229,
      "step": 500
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.7536793183578622,
      "eval_f1_macro": 0.46141164864389705,
      "eval_f1_micro": 0.7536793183578622,
      "eval_f1_weighted": 0.7337325056665881,
      "eval_loss": 0.7034952640533447,
      "eval_macro_fpr": 0.023326126599014015,
      "eval_macro_sensitivity": 0.4865735295815666,
      "eval_macro_specificity": 0.9815978320880049,
      "eval_precision": 0.7340611054176746,
      "eval_precision_macro": 0.4982390488816621,
      "eval_recall": 0.7536793183578622,
      "eval_recall_macro": 0.4865735295815666,
      "eval_runtime": 33.282,
      "eval_samples_per_second": 38.79,
      "eval_steps_per_second": 4.868,
      "eval_weighted_fpr": 0.022812051649928264,
      "eval_weighted_sensitivity": 0.7536793183578622,
      "eval_weighted_specificity": 0.9702881629622103,
      "step": 643
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.7652982184353214,
      "eval_f1_macro": 0.4824234674423381,
      "eval_f1_micro": 0.7652982184353214,
      "eval_f1_weighted": 0.7415202198150465,
      "eval_loss": 0.6963701844215393,
      "eval_macro_fpr": 0.022125198961045537,
      "eval_macro_sensitivity": 0.5099998711032655,
      "eval_macro_specificity": 0.982427621700107,
      "eval_precision": 0.7403884847240659,
      "eval_precision_macro": 0.5015009503316971,
      "eval_recall": 0.7652982184353214,
      "eval_recall_macro": 0.5099998711032655,
      "eval_runtime": 29.8835,
      "eval_samples_per_second": 43.201,
      "eval_steps_per_second": 5.421,
      "eval_weighted_fpr": 0.02143615139724089,
      "eval_weighted_sensitivity": 0.7652982184353214,
      "eval_weighted_specificity": 0.9711161070662828,
      "step": 803
    }
  ],
  "logging_steps": 500,
  "max_steps": 2400,
  "num_train_epochs": 15,
  "save_steps": 500,
  "total_flos": 6972431086817280.0,
  "trial_name": null,
  "trial_params": null
}