<!-- 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-expression_epoch5
This model is a fine-tuned version of bert-base-multilingual-cased on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5897
- Precision: 0.5835
- Recall: 0.5688
- F1: 0.5760
- Accuracy: 0.8344
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 | 218 | 0.5185 | 0.5076 | 0.5034 | 0.5055 | 0.8207 |
| No log | 2.0 | 436 | 0.4972 | 0.4948 | 0.5638 | 0.5271 | 0.8177 |
| 0.5193 | 3.0 | 654 | 0.5128 | 0.5838 | 0.5554 | 0.5692 | 0.8390 |
| 0.5193 | 4.0 | 872 | 0.5665 | 0.5612 | 0.6074 | 0.5834 | 0.8224 |
| 0.2063 | 5.0 | 1090 | 0.5897 | 0.5835 | 0.5688 | 0.5760 | 0.8344 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1+cu113
- Datasets 2.7.0
- Tokenizers 0.13.2
