EYEDOL/whisper-tiny-hausa2
This model is a fine-tuned version of EYEDOL/whisper-tiny-hausa1 on the EYEDOL/naija-voices-hausa-split_0-5 dataset. It achieves the following results on the evaluation set:
- Loss: 0.6946
- Wer Ortho: 0.5454
- Wer: 0.4809
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: 32
- eval_batch_size: 16
- 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: constant_with_warmup
- lr_scheduler_warmup_steps: 0.1
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 1.2967 | 1.0 | 665 | 0.6307 | 0.5191 | 0.4636 |
| 1.1962 | 2.0 | 1330 | 0.6195 | 0.5019 | 0.4473 |
| 1.0667 | 3.0 | 1995 | 0.6200 | 0.5036 | 0.4456 |
| 0.9621 | 4.0 | 2660 | 0.6227 | 0.5083 | 0.4455 |
| 0.8693 | 5.0 | 3325 | 0.6323 | 0.5126 | 0.4540 |
| 0.7838 | 6.0 | 3990 | 0.6426 | 0.5192 | 0.4556 |
| 0.7056 | 7.0 | 4655 | 0.6494 | 0.5218 | 0.4650 |
| 0.6303 | 8.0 | 5320 | 0.6652 | 0.5369 | 0.4758 |
| 0.5595 | 9.0 | 5985 | 0.6766 | 0.5332 | 0.4736 |
| 0.4927 | 10.0 | 6650 | 0.6946 | 0.5454 | 0.4809 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2