bert-base-uncased-finetuned-squad_v2
This model is a fine-tuned version of bert-base-uncased on the SQuAD2.0 dataset. It's been trained on question-answer pairs, including unanswerable questions, for the task of Question Answering.
It achieves the following results on the evaluation set:
- Loss: 1.7075
- Exact Match: 71.6920
- F1-score: 75.4437
Overview
Language model: bert-base-uncased
Language: English
Downstream-task: Extractive QA
Training data: SQuAD 2.0
Eval data: SQuAD 2.0
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: 5
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.0593 | 1.0 | 8235 | 1.1296 |
| 0.7736 | 2.0 | 16470 | 1.1290 |
| 0.5682 | 3.0 | 24705 | 1.1725 |
| 0.4124 | 4.0 | 32940 | 1.4632 |
| 0.3137 | 5.0 | 41175 | 1.7075 |
Framework versions
- Transformers 4.34.0
- Pytorch 1.12.1
- Datasets 2.14.5
- Tokenizers 0.14.1