<!-- 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. -->
hing-roberta-finetuned-ours-DS
This model is a fine-tuned version of l3cube-pune/hing-roberta on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.7267
- Accuracy: 0.68
- Precision: 0.6320
- Recall: 0.6234
- F1: 0.6133
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: 1.0638650088808569e-05
- train_batch_size: 16
- eval_batch_size: 32
- seed: 43
- 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 | Accuracy | Precision | Recall | F1 |
---|---|---|---|---|---|---|---|
1.0144 | 0.99 | 99 | 0.8326 | 0.585 | 0.4260 | 0.5589 | 0.4470 |
0.7626 | 1.98 | 198 | 0.7314 | 0.625 | 0.5366 | 0.5541 | 0.4828 |
0.6243 | 2.97 | 297 | 0.7064 | 0.65 | 0.5702 | 0.5888 | 0.5562 |
0.5539 | 3.96 | 396 | 0.7261 | 0.695 | 0.6598 | 0.6362 | 0.6326 |
0.4881 | 4.95 | 495 | 0.7267 | 0.68 | 0.6320 | 0.6234 | 0.6133 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.10.1+cu111
- Datasets 2.3.2
- Tokenizers 0.12.1