<!-- 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. -->
my_ner_model1_2_split_by_sentence_
This model is a fine-tuned version of dbmdz/bert-base-turkish-cased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.2257
- Precision: 0.6377
- Recall: 0.7644
- F1: 0.6953
- Accuracy: 0.9401
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: linear
- num_epochs: 5
Training results
Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|
No log | 1.0 | 427 | 0.2342 | 0.5265 | 0.6923 | 0.5981 | 0.9264 |
0.3433 | 2.0 | 854 | 0.1959 | 0.6024 | 0.7397 | 0.6640 | 0.9364 |
0.1686 | 3.0 | 1281 | 0.2085 | 0.6068 | 0.7608 | 0.6752 | 0.9376 |
0.1241 | 4.0 | 1708 | 0.2198 | 0.6376 | 0.7591 | 0.6931 | 0.9398 |
0.0912 | 5.0 | 2135 | 0.2257 | 0.6377 | 0.7644 | 0.6953 | 0.9401 |
Framework versions
- Transformers 4.22.1
- Pytorch 1.12.1+cu102
- Datasets 2.7.1
- Tokenizers 0.12.1