truecaser-lin-sna
This model is a fine-tuned version of Davlan/afro-xlmr-large-114L on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0276
- Accuracy: 0.9932
- Recall Lower: 0.9984
- Prec Lower: 0.9944
- Recall Cap: 0.9374
- Prec Cap: 0.9788
- Recall Upper: 0.0
- Prec Upper: 0.0
- Cap Recall: 0.9319
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: 32
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 5.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Recall Lower | Prec Lower | Recall Cap | Prec Cap | Recall Upper | Prec Upper | Cap Recall |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0611 | 0.1037 | 200 | 0.0382 | 0.9926 | 0.9973 | 0.9948 | 0.9418 | 0.9657 | 0.0 | 0.0 | 0.9362 |
| 0.0345 | 0.2074 | 400 | 0.0328 | 0.9920 | 0.9948 | 0.9970 | 0.9646 | 0.9348 | 0.0 | 0.0 | 0.9589 |
| 0.0303 | 0.3110 | 600 | 0.0285 | 0.9927 | 0.9984 | 0.9940 | 0.9305 | 0.9760 | 0.0 | 0.0 | 0.9250 |
| 0.033 | 0.4147 | 800 | 0.0272 | 0.9934 | 0.9974 | 0.9959 | 0.9499 | 0.9660 | 0.2105 | 0.3077 | 0.9456 |
| 0.0296 | 0.5184 | 1000 | 0.0270 | 0.9934 | 0.9965 | 0.9967 | 0.9609 | 0.9543 | 0.0526 | 1.0 | 0.9555 |
| 0.0327 | 0.6221 | 1200 | 0.0276 | 0.9932 | 0.9984 | 0.9944 | 0.9374 | 0.9788 | 0.0 | 0.0 | 0.9319 |
Framework versions
- Transformers 4.57.6
- Pytorch 2.13.0+cu130
- Datasets 3.6.0
- Tokenizers 0.22.2