funnel-transformer-xlarge_ner_conll2003
This model is a fine-tuned version of funnel-transformer/xlarge on the conll2003 dataset. It achieves the following results on the evaluation set:
- Loss: 0.0436
- Precision: 0.9565
- Recall: 0.9593
- F1: 0.9579
- Accuracy: 0.9915
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.1349 | 1.0 | 878 | 0.0441 | 0.9328 | 0.9438 | 0.9383 | 0.9881 |
| 0.0308 | 2.0 | 1756 | 0.0377 | 0.9457 | 0.9561 | 0.9509 | 0.9901 |
| 0.0144 | 3.0 | 2634 | 0.0432 | 0.9512 | 0.9578 | 0.9545 | 0.9906 |
| 0.007 | 4.0 | 3512 | 0.0419 | 0.9551 | 0.9584 | 0.9567 | 0.9913 |
| 0.0041 | 5.0 | 4390 | 0.0436 | 0.9565 | 0.9593 | 0.9579 | 0.9915 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.11.0
- Datasets 2.1.0
- Tokenizers 0.12.1