albert-base-v2
This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
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: 4
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|
| 0.024 | 1.0 | 4000 | 0.0300 |
| 0.0049 | 2.0 | 8000 | 0.0075 |
| 0.0 | 3.0 | 12000 | 0.0125 |
| 0.0 | 4.0 | 16000 | 0.0101 |
| 0.0056 | 5.0 | 20000 | 0.0104 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.9.0
- Datasets 2.1.0
- Tokenizers 0.12.1
