training_checkpoints/checkpoint-1344/trainer_state.json

{
  "best_metric": 0.6690246389304302,
  "best_model_checkpoint": "case-analysis-InLegalBERT/checkpoint-1344",
  "epoch": 6.0,
  "eval_steps": 500,
  "global_step": 1344,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8017817371937639,
      "eval_f1_macro": 0.5936102522033613,
      "eval_f1_micro": 0.8017817371937639,
      "eval_f1_weighted": 0.7819923339877444,
      "eval_loss": 0.6545999050140381,
      "eval_macro_fpr": 0.09781883735656886,
      "eval_macro_sensitivity": 0.6106003826592061,
      "eval_macro_specificity": 0.9112420642353884,
      "eval_precision": 0.763240968439436,
      "eval_precision_macro": 0.5776616905319997,
      "eval_recall": 0.8017817371937639,
      "eval_recall_macro": 0.6106003826592061,
      "eval_runtime": 9.2925,
      "eval_samples_per_second": 48.318,
      "eval_steps_per_second": 6.134,
      "eval_weighted_fpr": 0.07613344739093242,
      "eval_weighted_sensitivity": 0.8017817371937639,
      "eval_weighted_specificity": 0.84318651974779,
      "step": 224
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8129175946547884,
      "eval_f1_macro": 0.5995597405020452,
      "eval_f1_micro": 0.8129175946547884,
      "eval_f1_weighted": 0.7925821933419933,
      "eval_loss": 0.6830806732177734,
      "eval_macro_fpr": 0.0922965769071398,
      "eval_macro_sensitivity": 0.615426637485461,
      "eval_macro_specificity": 0.9170902440637451,
      "eval_precision": 0.7732439444566204,
      "eval_precision_macro": 0.5844969763606377,
      "eval_recall": 0.8129175946547884,
      "eval_recall_macro": 0.615426637485461,
      "eval_runtime": 36.5377,
      "eval_samples_per_second": 12.289,
      "eval_steps_per_second": 1.56,
      "eval_weighted_fpr": 0.07124681933842239,
      "eval_weighted_sensitivity": 0.8129175946547884,
      "eval_weighted_specificity": 0.8554433816001921,
      "step": 448
    },
    {
      "epoch": 2.23,
      "grad_norm": 6.910687446594238,
      "learning_rate": 4.6302083333333335e-05,
      "loss": 0.607,
      "step": 500
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8262806236080178,
      "eval_f1_macro": 0.6362135046138226,
      "eval_f1_micro": 0.8262806236080178,
      "eval_f1_weighted": 0.8105096248986687,
      "eval_loss": 0.7626442909240723,
      "eval_macro_fpr": 0.0885057964533289,
      "eval_macro_sensitivity": 0.6341414944356121,
      "eval_macro_specificity": 0.9181782637863818,
      "eval_precision": 0.8060163824047696,
      "eval_precision_macro": 0.6772816989326722,
      "eval_recall": 0.8262806236080178,
      "eval_recall_macro": 0.6341414944356121,
      "eval_runtime": 35.6726,
      "eval_samples_per_second": 12.587,
      "eval_steps_per_second": 1.598,
      "eval_weighted_fpr": 0.0654911838790932,
      "eval_weighted_sensitivity": 0.8262806236080178,
      "eval_weighted_specificity": 0.8464324315375095,
      "step": 672
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8084632516703786,
      "eval_f1_macro": 0.6313839669673806,
      "eval_f1_micro": 0.8084632516703786,
      "eval_f1_weighted": 0.8017481865377005,
      "eval_loss": 0.7839008569717407,
      "eval_macro_fpr": 0.08961602877899306,
      "eval_macro_sensitivity": 0.630579873226932,
      "eval_macro_specificity": 0.9209618973154744,
      "eval_precision": 0.7990879202460494,
      "eval_precision_macro": 0.6390850934968582,
      "eval_recall": 0.8084632516703786,
      "eval_recall_macro": 0.630579873226932,
      "eval_runtime": 33.6763,
      "eval_samples_per_second": 13.333,
      "eval_steps_per_second": 1.693,
      "eval_weighted_fpr": 0.07319148936170213,
      "eval_weighted_sensitivity": 0.8084632516703786,
      "eval_weighted_specificity": 0.8753843375915193,
      "step": 896
    },
    {
      "epoch": 4.46,
      "grad_norm": 0.3201627731323242,
      "learning_rate": 4.258184523809524e-05,
      "loss": 0.316,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8262806236080178,
      "eval_f1_macro": 0.6513945578231293,
      "eval_f1_micro": 0.8262806236080178,
      "eval_f1_weighted": 0.8161065406117904,
      "eval_loss": 0.9380576610565186,
      "eval_macro_fpr": 0.08224137352531583,
      "eval_macro_sensitivity": 0.6573161205514146,
      "eval_macro_specificity": 0.9260657452373569,
      "eval_precision": 0.8127349298285333,
      "eval_precision_macro": 0.668810150869688,
      "eval_recall": 0.8262806236080178,
      "eval_recall_macro": 0.6573161205514146,
      "eval_runtime": 34.3275,
      "eval_samples_per_second": 13.08,
      "eval_steps_per_second": 1.66,
      "eval_weighted_fpr": 0.0654911838790932,
      "eval_weighted_sensitivity": 0.8262806236080178,
      "eval_weighted_specificity": 0.8779823573414097,
      "step": 1120
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.821826280623608,
      "eval_f1_macro": 0.6690246389304302,
      "eval_f1_micro": 0.821826280623608,
      "eval_f1_weighted": 0.8159237291984361,
      "eval_loss": 1.0433590412139893,
      "eval_macro_fpr": 0.08967976675610558,
      "eval_macro_sensitivity": 0.6533037047742931,
      "eval_macro_specificity": 0.9186687903009794,
      "eval_precision": 0.8145202780918083,
      "eval_precision_macro": 0.6906556062141168,
      "eval_recall": 0.821826280623608,
      "eval_recall_macro": 0.6533037047742931,
      "eval_runtime": 43.316,
      "eval_samples_per_second": 10.366,
      "eval_steps_per_second": 1.316,
      "eval_weighted_fpr": 0.06739679865206402,
      "eval_weighted_sensitivity": 0.821826280623608,
      "eval_weighted_specificity": 0.8528488805803094,
      "step": 1344
    }
  ],
  "logging_steps": 500,
  "max_steps": 6720,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 30,
  "save_steps": 500,
  "total_flos": 2829020867592192.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}