sa_BERT_48_mnli
This model is a fine-tuned version of gokuls/bert_base_48 on the GLUE MNLI dataset. It achieves the following results on the evaluation set:
- Loss: 0.7082
- Accuracy: 0.7034
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: 4e-05
- train_batch_size: 96
- eval_batch_size: 96
- seed: 10
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 50
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 0.9145 | 1.0 | 4091 | 0.8006 | 0.6536 |
| 0.7442 | 2.0 | 8182 | 0.7245 | 0.6903 |
| 0.6631 | 3.0 | 12273 | 0.7323 | 0.6979 |
| 0.5942 | 4.0 | 16364 | 0.7073 | 0.7076 |
| 0.5241 | 5.0 | 20455 | 0.7475 | 0.7016 |
| 0.4526 | 6.0 | 24546 | 0.8377 | 0.7088 |
| 0.3842 | 7.0 | 28637 | 0.8736 | 0.6956 |
| 0.3213 | 8.0 | 32728 | 0.9334 | 0.6945 |
| 0.2669 | 9.0 | 36819 | 1.0196 | 0.7027 |
Framework versions
- Transformers 4.30.2
- Pytorch 1.14.0a0+410ce96
- Datasets 2.13.0
- Tokenizers 0.13.3