Edit model card

ECHR_test_2 Task A

This model is a fine-tuned version of nlpaueb/legal-bert-base-uncased on the lex_glue dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1998
  • Macro-f1: 0.5295
  • Micro-f1: 0.6157

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: 3e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Macro-f1 Micro-f1
0.2142 0.44 500 0.2887 0.2391 0.4263
0.172 0.89 1000 0.2672 0.2908 0.4628
0.1737 1.33 1500 0.2612 0.3657 0.5102
0.1581 1.78 2000 0.2412 0.3958 0.5468
0.1509 2.22 2500 0.2264 0.3950 0.5552
0.1606 2.67 3000 0.2342 0.4006 0.5511
0.1491 3.11 3500 0.2176 0.4558 0.5622
0.1392 3.56 4000 0.2454 0.4128 0.5596
0.15 4.0 4500 0.2113 0.4684 0.5874
0.1461 4.44 5000 0.2179 0.4631 0.5815
0.1457 4.89 5500 0.2151 0.4805 0.5949
0.1443 5.33 6000 0.2155 0.5123 0.5917
0.1279 5.78 6500 0.2131 0.4915 0.5998
0.1377 6.22 7000 0.2244 0.4705 0.5944
0.1242 6.67 7500 0.2150 0.5089 0.5918
0.1222 7.11 8000 0.2045 0.4801 0.5981
0.1372 7.56 8500 0.2074 0.5317 0.5962
0.1289 8.0 9000 0.2035 0.5323 0.6126
0.1295 8.44 9500 0.2058 0.5213 0.6073
0.123 8.89 10000 0.2027 0.5486 0.6135
0.1335 9.33 10500 0.1984 0.5442 0.6249
0.1258 9.78 11000 0.1998 0.5295 0.6157

Framework versions

  • Transformers 4.19.2
  • Pytorch 1.11.0+cu113
  • Datasets 2.2.2
  • Tokenizers 0.12.1
Downloads last month
16
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train QuentinKemperino/ECHR_test_2