RuleBert-v0.1-k0 / README.md
ribesstefano's picture
Initial version
e160645
metadata
license: mit
base_model: classla/xlm-roberta-base-multilingual-text-genre-classifier
tags:
  - Italian
  - legal ruling
  - generated_from_trainer
metrics:
  - f1
  - accuracy
model-index:
  - name: ribesstefano/RuleBert-v0.1-k0
    results: []

ribesstefano/RuleBert-v0.1-k0

This model is a fine-tuned version of classla/xlm-roberta-base-multilingual-text-genre-classifier on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3765
  • F1: 0.5087
  • Roc Auc: 0.6757
  • Accuracy: 0.0333

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 4
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • training_steps: 4000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss F1 Roc Auc Accuracy
0.3621 0.14 250 0.3791 0.4861 0.6678 0.0458
0.3199 0.28 500 0.3754 0.4969 0.6730 0.0375
0.3135 0.41 750 0.3785 0.4965 0.6715 0.0458
0.3176 0.55 1000 0.3755 0.4988 0.6725 0.0333
0.3056 0.69 1250 0.3734 0.5100 0.6756 0.0667
0.3016 0.83 1500 0.3749 0.5076 0.6755 0.0333
0.3044 0.97 1750 0.3763 0.5014 0.6738 0.0125
0.3009 1.11 2000 0.3781 0.5073 0.6751 0.0333
0.3044 1.24 2250 0.3782 0.5089 0.6755 0.0417
0.2979 1.38 2500 0.3770 0.5047 0.6745 0.0167
0.2977 1.52 2750 0.3772 0.5114 0.6764 0.0417
0.2998 1.66 3000 0.3747 0.5107 0.6762 0.0375
0.2916 1.8 3250 0.3741 0.5096 0.6758 0.0375
0.3107 1.94 3500 0.3757 0.5056 0.6748 0.0208
0.3023 2.07 3750 0.3767 0.5069 0.6752 0.025
0.3069 2.21 4000 0.3765 0.5087 0.6757 0.0333

Framework versions

  • Transformers 4.36.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0