distilbert-base-uncased-test2
This model is a fine-tuned version of distilbert-base-uncased on the wnut_17 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3055
- Precision: 0.5278
- Recall: 0.3957
- F1: 0.4523
- Accuracy: 0.9462
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|
No log | 1.0 | 213 | 0.2889 | 0.5439 | 0.3503 | 0.4262 | 0.9453 |
No log | 2.0 | 426 | 0.2938 | 0.5236 | 0.3800 | 0.4404 | 0.9457 |
0.0544 | 3.0 | 639 | 0.3055 | 0.5278 | 0.3957 | 0.4523 | 0.9462 |
Framework versions
- Transformers 4.21.1
- Pytorch 1.12.1+cu102
- Datasets 2.4.0
- Tokenizers 0.12.1
- Downloads last month
- 113
Inference Providers
NEW
This model is not currently available via any of the supported Inference Providers.
Dataset used to train Jethuestad/distilbert-base-uncased-test2
Evaluation results
- Precision on wnut_17self-reported0.528
- Recall on wnut_17self-reported0.396
- F1 on wnut_17self-reported0.452
- Accuracy on wnut_17self-reported0.946