sunbird-whisper-lin-sna-V2
This model is a fine-tuned version of Sunbird/asr-whisper-51-african-languages on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1994
- Wer: 0.2508
- Cer: 0.0425
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: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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_steps: 200
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.5210 | 0.2770 | 200 | 0.1919 | 0.2417 | 0.0410 |
| 0.5012 | 0.5540 | 400 | 0.1990 | 0.2535 | 0.0425 |
| 0.5299 | 0.8310 | 600 | 0.1958 | 0.2484 | 0.0423 |
| 0.3421 | 1.1080 | 800 | 0.1994 | 0.2508 | 0.0425 |
Framework versions
- Transformers 5.14.1
- Pytorch 2.11.0+cu130
- Datasets 5.0.1
- Tokenizers 0.22.2