--- Part of BA Thesis by Enis Settouf @ HTW Berlin Business computing ---
This model was trained additionally on a MLM task for the legal domain on German laws and legal Cases
Data Source: OpenLegalData.io
The rest of this model card has been generated automatically:
<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
cross-en-de-roberta-sentence-transformer-openlegal
This model is a fine-tuned version of T-Systems-onsite/cross-en-de-roberta-sentence-transformer on the None dataset. It achieves the following results on the evaluation set:
- Loss: 2.3120
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: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: tpu
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 6.4835 | 0.06 | 500 | 5.2259 |
| 4.9569 | 0.13 | 1000 | 4.5163 |
| 4.4158 | 0.19 | 1500 | 4.1257 |
| 4.1221 | 0.25 | 2000 | 3.8695 |
| 3.8853 | 0.32 | 2500 | 3.6586 |
| 3.7092 | 0.38 | 3000 | 3.5126 |
| 3.5779 | 0.45 | 3500 | 3.3889 |
| 3.4424 | 0.51 | 4000 | 3.2731 |
| 3.3556 | 0.57 | 4500 | 3.1984 |
| 3.2627 | 0.64 | 5000 | 3.1103 |
| 3.1855 | 0.7 | 5500 | 3.0306 |
| 3.1381 | 0.76 | 6000 | 2.9796 |
| 3.0763 | 0.83 | 6500 | 2.9299 |
| 2.9985 | 0.89 | 7000 | 2.8740 |
| 2.9359 | 0.95 | 7500 | 2.8300 |
| 2.8954 | 1.02 | 8000 | 2.7861 |
| 2.8322 | 1.08 | 8500 | 2.7450 |
| 2.816 | 1.14 | 9000 | 2.7255 |
| 2.8013 | 1.21 | 9500 | 2.6872 |
| 2.7414 | 1.27 | 10000 | 2.6538 |
| 2.707 | 1.34 | 10500 | 2.6284 |
| 2.6866 | 1.4 | 11000 | 2.6021 |
| 2.6429 | 1.46 | 11500 | 2.5721 |
| 2.6269 | 1.53 | 12000 | 2.5646 |
| 2.6173 | 1.59 | 12500 | 2.5323 |
| 2.5959 | 1.65 | 13000 | 2.5052 |
| 2.5692 | 1.72 | 13500 | 2.4993 |
| 2.5563 | 1.78 | 14000 | 2.4840 |
| 2.5448 | 1.84 | 14500 | 2.4635 |
| 2.4932 | 1.91 | 15000 | 2.4581 |
| 2.5106 | 1.97 | 15500 | 2.4342 |
| 2.5009 | 2.03 | 16000 | 2.4260 |
| 2.46 | 2.1 | 16500 | 2.4152 |
| 2.4417 | 2.16 | 17000 | 2.4079 |
| 2.4568 | 2.23 | 17500 | 2.4010 |
| 2.442 | 2.29 | 18000 | 2.3875 |
| 2.4328 | 2.35 | 18500 | 2.3724 |
| 2.4126 | 2.42 | 19000 | 2.3645 |
| 2.4063 | 2.48 | 19500 | 2.3612 |
| 2.362 | 2.54 | 20000 | 2.3565 |
| 2.3877 | 2.61 | 20500 | 2.3507 |
| 2.3839 | 2.67 | 21000 | 2.3353 |
| 2.3657 | 2.73 | 21500 | 2.3326 |
| 2.3464 | 2.8 | 22000 | 2.3262 |
| 2.3915 | 2.86 | 22500 | 2.3259 |
| 2.3613 | 2.93 | 23000 | 2.3195 |
| 2.358 | 2.99 | 23500 | 2.3165 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.8.2+cpu
- Datasets 2.4.0
- Tokenizers 0.12.1
