<!-- 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-de-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: 0.3820
- Precision: 0.5214
- Recall: 0.5660
- F1: 0.5428
- Accuracy: 0.8966
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 |
---|---|---|---|---|---|---|---|
No log | 1.0 | 463 | 0.6140 | 0.2884 | 0.2925 | 0.2904 | 0.8438 |
0.8329 | 2.0 | 926 | 0.4504 | 0.4092 | 0.4423 | 0.4251 | 0.8720 |
0.4385 | 3.0 | 1389 | 0.4046 | 0.4634 | 0.5042 | 0.4829 | 0.8875 |
0.3364 | 4.0 | 1852 | 0.3843 | 0.5 | 0.5446 | 0.5213 | 0.8954 |
0.2919 | 5.0 | 2315 | 0.3820 | 0.5214 | 0.5660 | 0.5428 | 0.8966 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.1+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2