Whisper Small Luganda
This model is a fine-tuned version of openai/whisper-small on the Common Voice 15.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3827
- Wer: 40.4482
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.6714 | 0.11 | 500 | 0.7162 | 66.8818 |
0.4726 | 0.23 | 1000 | 0.5434 | 54.5155 |
0.4208 | 0.34 | 1500 | 0.4767 | 49.1337 |
0.3882 | 0.45 | 2000 | 0.4404 | 45.2067 |
0.3736 | 0.56 | 2500 | 0.4166 | 44.0255 |
0.3387 | 0.68 | 3000 | 0.3994 | 41.2638 |
0.3403 | 0.79 | 3500 | 0.3886 | 41.0884 |
0.3088 | 0.9 | 4000 | 0.3827 | 40.4482 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 1
Finetuned from
Dataset used to train mn720/english
Evaluation results
- Wer on Common Voice 15.0validation set self-reported40.448