Sunbird-2-sunbird-gikuyu-financial
This model is a fine-tuned version of Sunbird/asr-whisper-51-african-languages on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.1145
- Wer: 64.1834
- Cer: 43.5638
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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- training_steps: 400
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.9103 | 0.9615 | 50 | 0.7997 | 58.3557 | 41.3460 |
| 0.622 | 1.9231 | 100 | 0.7425 | 57.1365 | 41.3181 |
| 0.5328 | 2.8846 | 150 | 0.7579 | 59.1611 | 42.0318 |
| 0.3886 | 3.8462 | 200 | 0.7943 | 60.1119 | 42.0824 |
| 0.2601 | 4.8077 | 250 | 0.8768 | 60.6823 | 42.5518 |
| 0.1694 | 5.7692 | 300 | 0.9684 | 61.6331 | 42.8990 |
| 0.1106 | 6.7308 | 350 | 1.0621 | 64.5190 | 43.3353 |
| 0.0825 | 7.6923 | 400 | 1.1145 | 64.1834 | 43.5638 |
Framework versions
- Transformers 4.49.0
- Pytorch 2.7.0a0+79aa17489c.nv25.04
- Datasets 2.21.0
- Tokenizers 0.21.4