<!-- 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. -->
xlm-roberta-base-uk-aug-ner
This model is a fine-tuned version of xlm-roberta-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.2396
- Precision: 0.6410
- Recall: 0.6370
- F1: 0.6390
- Accuracy: 0.9313
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: 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: 5
Training results
Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|
0.6689 | 1.0 | 725 | 0.3466 | 0.4923 | 0.4758 | 0.4839 | 0.9054 |
0.3286 | 2.0 | 1450 | 0.2742 | 0.5723 | 0.5894 | 0.5807 | 0.9179 |
0.2224 | 3.0 | 2175 | 0.2534 | 0.5884 | 0.6247 | 0.6060 | 0.9240 |
0.2008 | 4.0 | 2900 | 0.2396 | 0.6410 | 0.6370 | 0.6390 | 0.9313 |
0.1703 | 5.0 | 3625 | 0.2398 | 0.6336 | 0.6476 | 0.6405 | 0.9306 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.1+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2