<!-- 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-cola-v6
This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4210
- Accuracy: 0.9310
- Precision: 0.9310
- Recall: 0.9310
- F1: 0.9310
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: 30
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
---|---|---|---|---|---|---|---|
No log | 6.25 | 50 | 1.2900 | 0.5862 | 0.5862 | 0.5862 | 0.5862 |
No log | 12.5 | 100 | 0.6947 | 0.8621 | 0.8621 | 0.8621 | 0.8621 |
No log | 18.75 | 150 | 0.4672 | 0.9310 | 0.9310 | 0.9310 | 0.9310 |
No log | 25.0 | 200 | 0.4265 | 0.9310 | 0.9310 | 0.9310 | 0.9310 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1+cu113
- Datasets 2.7.1
- Tokenizers 0.13.2