HsscBERT_abs_and_full
This model is a fine-tuned version of /home/hscrc/pretrained_models/bert-base-chinese on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6037
- Accuracy: 0.8504
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: 32
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 9
- total_train_batch_size: 288
- total_eval_batch_size: 144
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5.0
Training results
| Training Loss | Epoch | Step | Accuracy | Validation Loss |
|---|---|---|---|---|
| 0.8163 | 0.19 | 5000 | 0.8326 | 0.6971 |
| 0.7942 | 0.38 | 10000 | 0.8364 | 0.6761 |
| 0.7817 | 0.57 | 15000 | 0.8384 | 0.6651 |
| 0.7751 | 0.75 | 20000 | 0.8402 | 0.6563 |
| 0.7654 | 0.94 | 25000 | 0.8415 | 0.6490 |
| 0.7546 | 1.13 | 30000 | 0.8427 | 0.6441 |
| 0.7527 | 1.32 | 35000 | 0.8434 | 0.6398 |
| 0.7484 | 1.51 | 40000 | 0.8444 | 0.6345 |
| 0.7443 | 1.7 | 45000 | 0.8450 | 0.6318 |
| 0.74 | 1.88 | 50000 | 0.8456 | 0.6292 |
| 0.738 | 2.07 | 55000 | 0.8460 | 0.6268 |
| 0.734 | 2.26 | 60000 | 0.8464 | 0.6246 |
| 0.7335 | 2.45 | 65000 | 0.8467 | 0.6229 |
| 0.7299 | 2.64 | 70000 | 0.8470 | 0.6212 |
| 0.7291 | 2.83 | 75000 | 0.8473 | 0.6201 |
Framework versions
- Transformers 4.26.1
- Pytorch 1.10.0+cu113
- Datasets 2.9.0
- Tokenizers 0.13.2