chinese-bert-wwm-finetuned-chnsenticorp
This model is a fine-tuned version of hfl/chinese-bert-wwm on a small subset of chnsenticorp dataset. It achieves the following results on the evaluation set:
- Loss: 3.0868
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: 5
- eval_batch_size: 5
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.0096 | 1.0 | 15 | 3.7742 |
| 1.7336 | 2.0 | 30 | 3.9102 |
| 2.5286 | 3.0 | 45 | 3.4744 |
| 2.8892 | 4.0 | 60 | 3.1142 |
| 2.7188 | 5.0 | 75 | 2.7622 |
| 2.7923 | 6.0 | 90 | 3.1119 |
| 2.4094 | 7.0 | 105 | 3.0426 |
| 2.5928 | 8.0 | 120 | 2.8928 |
| 2.4072 | 9.0 | 135 | 2.9462 |
| 2.4349 | 10.0 | 150 | 2.7645 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2