kobigbird-pure148-81233921
This model is a fine-tuned version of monologg/kobigbird-bert-base on the custom_squad_v2 dataset. It achieves the following results on the evaluation set:
- Loss: 1.0525
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: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 148
- gradient_accumulation_steps: 8
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| No log | 0.99 | 42 | 1.2940 |
| No log | 1.99 | 84 | 1.0463 |
| No log | 2.99 | 126 | 1.0525 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.1+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2