kobigbird-bert-base-finetuned-klue-v2_epoch64
This model is a fine-tuned version of monologg/kobigbird-bert-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.8139
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: 1e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 4
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.5026 | 0.55 | 500 | 1.3754 |
| 0.9551 | 1.1 | 1000 | 0.9633 |
| 0.8 | 1.64 | 1500 | 0.8191 |
| 0.6182 | 2.19 | 2000 | 0.7808 |
| 0.5731 | 2.74 | 2500 | 0.7482 |
| 0.3699 | 3.29 | 3000 | 0.8175 |
| 0.3634 | 3.84 | 3500 | 0.8139 |
Framework versions
- Transformers 4.23.0
- Pytorch 1.12.1+cu113
- Datasets 2.5.2
- Tokenizers 0.13.1