<!-- 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. -->
bert-finetuned-targetexpressionaug_epoch5
This model is a fine-tuned version of bert-base-multilingual-cased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.2460
- Precision: 0.6388
- Recall: 0.6574
- F1: 0.6480
- Accuracy: 0.7685
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
- 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 |
---|---|---|---|---|---|---|---|
No log | 1.0 | 424 | 0.9757 | 0.5661 | 0.6406 | 0.6010 | 0.7455 |
0.2463 | 2.0 | 848 | 1.0356 | 0.6151 | 0.6350 | 0.6249 | 0.7656 |
0.1394 | 3.0 | 1272 | 1.0995 | 0.6246 | 0.6406 | 0.6325 | 0.7634 |
0.1155 | 4.0 | 1696 | 1.1802 | 0.6331 | 0.6529 | 0.6429 | 0.7673 |
0.0768 | 5.0 | 2120 | 1.2460 | 0.6388 | 0.6574 | 0.6480 | 0.7685 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1+cu113
- Datasets 2.7.0
- Tokenizers 0.13.2