<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
distilbert-base-uncased-finetuned-sst2-finetuned-sst2
This model was trained from scratch on the glue dataset. It achieves the following results on the evaluation set:
- Loss: 0.3156
- Accuracy: 0.9071
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: 256
- eval_batch_size: 256
- 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 | Accuracy |
---|---|---|---|---|
No log | 1.0 | 264 | 0.2830 | 0.9014 |
0.111 | 2.0 | 528 | 0.3156 | 0.9071 |
0.111 | 3.0 | 792 | 0.3351 | 0.8979 |
0.0688 | 4.0 | 1056 | 0.3377 | 0.9037 |
0.0688 | 5.0 | 1320 | 0.3526 | 0.9048 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1
- Datasets 2.7.1
- Tokenizers 0.13.2