metadata
library_name: transformers
language:
- ha
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
datasets:
- EYEDOL/naija-voices-hausa-split_0-0
metrics:
- wer
model-index:
- name: EYEDOL/whisper-tiny-hausa
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: EYEDOL/naija-voices-hausa-split_0-0
type: EYEDOL/naija-voices-hausa-split_0-0
metrics:
- name: Wer
type: wer
value: 0.5221648941771085
EYEDOL/whisper-tiny-hausa
This model is a fine-tuned version of openai/whisper-tiny on the EYEDOL/naija-voices-hausa-split_0-0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.7752
- Wer Ortho: 0.5775
- Wer: 0.5222
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 |
|---|---|---|---|---|---|
| 4.5185 | 1.0 | 665 | 1.2291 | 0.8407 | 0.7963 |
| 2.0880 | 2.0 | 1330 | 0.9774 | 0.7245 | 0.6717 |
| 1.6905 | 3.0 | 1995 | 0.8828 | 0.6599 | 0.6016 |
| 1.4668 | 4.0 | 2660 | 0.8334 | 0.6431 | 0.5829 |
| 1.3050 | 5.0 | 3325 | 0.7984 | 0.6149 | 0.5562 |
| 1.1746 | 6.0 | 3990 | 0.7819 | 0.6115 | 0.5516 |
| 1.0632 | 7.0 | 4655 | 0.7707 | 0.5996 | 0.5419 |
| 0.9630 | 8.0 | 5320 | 0.7678 | 0.5939 | 0.5360 |
| 0.8731 | 9.0 | 5985 | 0.7667 | 0.5963 | 0.5376 |
| 0.7893 | 10.0 | 6650 | 0.7752 | 0.5775 | 0.5222 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2