distil-slovakbert-upos
This model is a fine-tuned version of crabz/distil-slovakbert on the universal_dependencies sk_snk dataset. It achieves the following results on the evaluation set:
- Loss: 0.1207
- Precision: 0.9771
- Recall: 0.9785
- F1: 0.9778
- Accuracy: 0.9801
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: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 266 | 0.2168 | 0.9570 | 0.9554 | 0.9562 | 0.9610 |
| 0.3935 | 2.0 | 532 | 0.1416 | 0.9723 | 0.9736 | 0.9730 | 0.9740 |
| 0.3935 | 3.0 | 798 | 0.1236 | 0.9722 | 0.9735 | 0.9728 | 0.9747 |
| 0.0664 | 4.0 | 1064 | 0.1195 | 0.9722 | 0.9741 | 0.9732 | 0.9766 |
| 0.0664 | 5.0 | 1330 | 0.1160 | 0.9764 | 0.9772 | 0.9768 | 0.9789 |
| 0.0377 | 6.0 | 1596 | 0.1194 | 0.9763 | 0.9776 | 0.9770 | 0.9790 |
| 0.0377 | 7.0 | 1862 | 0.1188 | 0.9740 | 0.9755 | 0.9748 | 0.9777 |
| 0.024 | 8.0 | 2128 | 0.1188 | 0.9762 | 0.9777 | 0.9769 | 0.9793 |
| 0.024 | 9.0 | 2394 | 0.1207 | 0.9774 | 0.9789 | 0.9781 | 0.9802 |
| 0.0184 | 10.0 | 2660 | 0.1207 | 0.9771 | 0.9785 | 0.9778 | 0.9801 |
Framework versions
- Transformers 4.17.0.dev0
- Pytorch 1.10.0
- Datasets 1.16.1
- Tokenizers 0.11.0