<!-- 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-v3
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: 0.9655
- Matthews Correlation: 0.7369
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: 15
Training results
| Training Loss | Epoch | Step | Validation Loss | Matthews Correlation |
|---|---|---|---|---|
| No log | 1.0 | 8 | 1.9112 | 0.1486 |
| No log | 2.0 | 16 | 1.8626 | 0.1273 |
| No log | 3.0 | 24 | 1.7793 | 0.1947 |
| No log | 4.0 | 32 | 1.6722 | 0.1681 |
| No log | 5.0 | 40 | 1.5578 | 0.3876 |
| No log | 6.0 | 48 | 1.4463 | 0.5551 |
| No log | 7.0 | 56 | 1.3280 | 0.5498 |
| No log | 8.0 | 64 | 1.2302 | 0.5936 |
| No log | 9.0 | 72 | 1.1408 | 0.6998 |
| No log | 10.0 | 80 | 1.0765 | 0.6601 |
| No log | 11.0 | 88 | 1.0145 | 0.6988 |
| No log | 12.0 | 96 | 0.9655 | 0.7369 |
| No log | 13.0 | 104 | 0.9389 | 0.6992 |
| No log | 14.0 | 112 | 0.9258 | 0.6992 |
| No log | 15.0 | 120 | 0.9209 | 0.6992 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1+cu113
- Datasets 2.7.1
- Tokenizers 0.13.2
