bert-base-chinese-wikiann-zh-ner-new
This model is a fine-tuned version of bert-base-chinese on the wikiann dataset. It achieves the following results on the evaluation set:
- Loss: 0.2085
- Precision: 0.7834
- Recall: 0.8069
- F1: 0.7950
- Accuracy: 0.9435
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
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.773 | 0.16 | 400 | 0.3232 | 0.5816 | 0.6670 | 0.6214 | 0.9049 |
| 0.3149 | 0.32 | 800 | 0.2954 | 0.6832 | 0.6923 | 0.6877 | 0.9195 |
| 0.2912 | 0.48 | 1200 | 0.2418 | 0.7010 | 0.7551 | 0.7270 | 0.9299 |
| 0.2446 | 0.64 | 1600 | 0.2539 | 0.7159 | 0.7743 | 0.7440 | 0.9292 |
| 0.2193 | 0.8 | 2000 | 0.2330 | 0.7441 | 0.7613 | 0.7526 | 0.9351 |
| 0.2434 | 0.96 | 2400 | 0.2186 | 0.7603 | 0.7696 | 0.7649 | 0.9369 |
| 0.1915 | 1.12 | 2800 | 0.2245 | 0.7568 | 0.8032 | 0.7793 | 0.9398 |
| 0.1607 | 1.28 | 3200 | 0.2263 | 0.7566 | 0.8138 | 0.7842 | 0.9399 |
| 0.1513 | 1.44 | 3600 | 0.2228 | 0.7782 | 0.7964 | 0.7872 | 0.9414 |
| 0.1777 | 1.6 | 4000 | 0.2098 | 0.7857 | 0.7916 | 0.7887 | 0.9423 |
| 0.1466 | 1.76 | 4400 | 0.2132 | 0.7673 | 0.8163 | 0.7911 | 0.9418 |
| 0.1528 | 1.92 | 4800 | 0.2093 | 0.7793 | 0.8114 | 0.7951 | 0.9435 |
Framework versions
- Transformers 4.27.4
- Pytorch 2.0.0+cu118
- Datasets 2.11.0
- Tokenizers 0.13.3