<!-- 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. -->
bkk-ner-model
This model is a fine-tuned version of Geotrend/bert-base-th-cased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0518
- Precision: 0.8850
- Recall: 0.9615
- F1: 0.9217
- Accuracy: 0.9822
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: 8
- 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 |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 8 | 0.5592 | 0.3698 | 0.6827 | 0.4797 | 0.7818 |
| No log | 2.0 | 16 | 0.4491 | 0.4831 | 0.8269 | 0.6099 | 0.8062 |
| No log | 3.0 | 24 | 0.3738 | 0.6226 | 0.9519 | 0.7529 | 0.8399 |
| No log | 4.0 | 32 | 0.1781 | 0.6691 | 0.8942 | 0.7654 | 0.9401 |
| No log | 5.0 | 40 | 0.2201 | 0.8095 | 0.9808 | 0.8870 | 0.9204 |
| No log | 6.0 | 48 | 0.0936 | 0.8130 | 0.9615 | 0.8811 | 0.9710 |
| No log | 7.0 | 56 | 0.0692 | 0.8197 | 0.9615 | 0.8850 | 0.9757 |
| No log | 8.0 | 64 | 0.0712 | 0.8264 | 0.9615 | 0.8889 | 0.9710 |
| No log | 9.0 | 72 | 0.0575 | 0.8621 | 0.9615 | 0.9091 | 0.9803 |
| No log | 10.0 | 80 | 0.0625 | 0.8487 | 0.9712 | 0.9058 | 0.9766 |
| No log | 11.0 | 88 | 0.0580 | 0.8584 | 0.9327 | 0.8940 | 0.9766 |
| No log | 12.0 | 96 | 0.0551 | 0.8684 | 0.9519 | 0.9083 | 0.9813 |
| No log | 13.0 | 104 | 0.0554 | 0.8761 | 0.9519 | 0.9124 | 0.9803 |
| No log | 14.0 | 112 | 0.0535 | 0.8772 | 0.9615 | 0.9174 | 0.9813 |
| No log | 15.0 | 120 | 0.0518 | 0.8850 | 0.9615 | 0.9217 | 0.9822 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
