tiny-bert-sst2-distilled
This model is a fine-tuned version of google/bert_uncased_L-2_H-128_A-2 on the glue dataset. It achieves the following results on the evaluation set:
- Loss: 1.7305
- Accuracy: 0.8326
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: 0.0007199555649276667
- train_batch_size: 1024
- eval_batch_size: 1024
- seed: 33
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 7
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.77 | 1.0 | 66 | 1.6939 | 0.8165 |
| 0.729 | 2.0 | 132 | 1.5090 | 0.8326 |
| 0.5242 | 3.0 | 198 | 1.5369 | 0.8257 |
| 0.4017 | 4.0 | 264 | 1.7025 | 0.8326 |
| 0.327 | 5.0 | 330 | 1.6743 | 0.8245 |
| 0.2749 | 6.0 | 396 | 1.7305 | 0.8337 |
| 0.2521 | 7.0 | 462 | 1.7305 | 0.8326 |
Framework versions
- Transformers 4.12.3
- Pytorch 1.9.1
- Datasets 1.15.1
- Tokenizers 0.10.3