aristo-roberta-finetuned-csqa
This model is a fine-tuned version of LIAMF-USP/aristo-roberta on the commonsense_qa dataset. It achieves the following results on the evaluation set:
- Loss: 1.2187
- Accuracy: 0.7305
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: 1e-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
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.131 | 1.0 | 609 | 0.7109 | 0.7232 |
| 0.6957 | 2.0 | 1218 | 0.6912 | 0.7346 |
| 0.459 | 3.0 | 1827 | 0.8364 | 0.7305 |
| 0.3063 | 4.0 | 2436 | 1.0595 | 0.7322 |
| 0.2283 | 5.0 | 3045 | 1.2187 | 0.7305 |
Framework versions
- Transformers 4.9.0
- Pytorch 1.9.0
- Datasets 1.10.2
- Tokenizers 0.10.3