tiny-bert-sst2-distilled
This model is a fine-tuned version of google/bert_uncased_L-2_H-128_A-2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.4425
- Accuracy: 0.8280
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: 0.0009145413682990444
- train_batch_size: 8192
- eval_batch_size: 8192
- seed: 33
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 8
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 2.8528 | 1.0 | 9 | 2.1211 | 0.7523 |
| 1.6717 | 2.0 | 18 | 1.6573 | 0.7947 |
| 1.0433 | 3.0 | 27 | 1.3459 | 0.8108 |
| 0.7426 | 4.0 | 36 | 1.3367 | 0.8222 |
| 0.6055 | 5.0 | 45 | 1.3740 | 0.8245 |
| 0.5311 | 6.0 | 54 | 1.4192 | 0.8303 |
| 0.4907 | 7.0 | 63 | 1.4403 | 0.8257 |
| 0.4664 | 8.0 | 72 | 1.4425 | 0.8280 |
Framework versions
- Transformers 4.26.0
- Pytorch 1.13.1+cu117
- Datasets 2.16.1
- Tokenizers 0.13.3