CNEC1_1_xlm-roberta-large
This model is a fine-tuned version of FacebookAI/xlm-roberta-large on the cnec dataset. It achieves the following results on the evaluation set:
- Loss: 0.3816
- Precision: 0.8521
- Recall: 0.8721
- F1: 0.8620
- Accuracy: 0.9513
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: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 15
Training results
Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|
0.4004 | 1.0 | 1174 | 0.2747 | 0.7598 | 0.7876 | 0.7735 | 0.9381 |
0.2765 | 2.0 | 2348 | 0.2268 | 0.8181 | 0.8340 | 0.8260 | 0.9506 |
0.2104 | 3.0 | 3522 | 0.2400 | 0.8318 | 0.8561 | 0.8438 | 0.9524 |
0.1713 | 4.0 | 4696 | 0.2285 | 0.8353 | 0.8645 | 0.8496 | 0.9552 |
0.1241 | 5.0 | 5870 | 0.2278 | 0.8458 | 0.8715 | 0.8584 | 0.9585 |
0.0997 | 6.0 | 7044 | 0.2717 | 0.8372 | 0.8653 | 0.8511 | 0.9559 |
0.0878 | 7.0 | 8218 | 0.2599 | 0.8439 | 0.8830 | 0.8630 | 0.9583 |
0.0585 | 8.0 | 9392 | 0.2868 | 0.8415 | 0.8764 | 0.8586 | 0.9564 |
0.0489 | 9.0 | 10566 | 0.2900 | 0.8594 | 0.8795 | 0.8693 | 0.9568 |
0.0416 | 10.0 | 11740 | 0.3061 | 0.8646 | 0.8852 | 0.8748 | 0.9598 |
0.0316 | 11.0 | 12914 | 0.3240 | 0.8567 | 0.8843 | 0.8703 | 0.9576 |
0.0264 | 12.0 | 14088 | 0.3329 | 0.8546 | 0.8795 | 0.8668 | 0.9588 |
0.0184 | 13.0 | 15262 | 0.3475 | 0.8628 | 0.8804 | 0.8715 | 0.9584 |
0.0156 | 14.0 | 16436 | 0.3472 | 0.8654 | 0.8826 | 0.8739 | 0.9592 |
0.0125 | 15.0 | 17610 | 0.3539 | 0.8670 | 0.8861 | 0.8764 | 0.9593 |
Framework versions
- Transformers 4.36.2
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0
- Downloads last month
- 9
Finetuned from
Evaluation results
- Precision on cnecvalidation set self-reported0.852
- Recall on cnecvalidation set self-reported0.872
- F1 on cnecvalidation set self-reported0.862
- Accuracy on cnecvalidation set self-reported0.951