tiny-bert-sst2-distilled-model
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: 1.2592
- Accuracy: 0.8383
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: 6e-05
- train_batch_size: 16
- eval_batch_size: 16
- 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 |
|---|---|---|---|---|
| 0.5303 | 1.0 | 4210 | 1.2542 | 0.8222 |
| 0.4503 | 2.0 | 8420 | 1.1260 | 0.8211 |
| 0.3689 | 3.0 | 12630 | 1.2325 | 0.8234 |
| 0.3122 | 4.0 | 16840 | 1.2533 | 0.8337 |
| 0.2764 | 5.0 | 21050 | 1.2726 | 0.8337 |
| 0.254 | 6.0 | 25260 | 1.2609 | 0.8337 |
| 0.2358 | 7.0 | 29470 | 1.2592 | 0.8383 |
Framework versions
- Transformers 4.19.2
- Pytorch 1.10.1+cu113
- Datasets 1.15.1
- Tokenizers 0.12.1