XLS-R-LUGANDA-ASR-CV14
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_14_0 dataset. It achieves the following results on the evaluation set:
- Loss: inf
- Wer: 0.2406
- Cer: 0.0537
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: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 10000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 4.24 | 0.18 | 400 | inf | 0.8354 | 0.2170 |
| 0.6124 | 0.36 | 800 | inf | 0.5690 | 0.1360 |
| 0.4411 | 0.54 | 1200 | inf | 0.4746 | 0.1120 |
| 0.3839 | 0.72 | 1600 | inf | 0.4409 | 0.1050 |
| 0.3504 | 0.9 | 2000 | inf | 0.3955 | 0.0943 |
| 0.3214 | 1.08 | 2400 | inf | 0.3678 | 0.0854 |
| 0.2879 | 1.26 | 2800 | inf | 0.3614 | 0.0836 |
| 0.284 | 1.45 | 3200 | inf | 0.3411 | 0.0789 |
| 0.2683 | 1.63 | 3600 | inf | 0.3362 | 0.0767 |
| 0.2572 | 1.81 | 4000 | inf | 0.3241 | 0.0740 |
| 0.2532 | 1.99 | 4400 | inf | 0.3117 | 0.0719 |
| 0.2228 | 2.17 | 4800 | inf | 0.2977 | 0.0677 |
| 0.2143 | 2.35 | 5200 | inf | 0.2969 | 0.0676 |
| 0.211 | 2.53 | 5600 | inf | 0.2918 | 0.0665 |
| 0.2066 | 2.71 | 6000 | inf | 0.2848 | 0.0647 |
| 0.2026 | 2.89 | 6400 | inf | 0.2804 | 0.0637 |
| 0.1898 | 3.07 | 6800 | inf | 0.2744 | 0.0627 |
| 0.1747 | 3.25 | 7200 | inf | 0.2668 | 0.0603 |
| 0.1667 | 3.43 | 7600 | inf | 0.2631 | 0.0597 |
| 0.1639 | 3.61 | 8000 | inf | 0.2558 | 0.0580 |
| 0.1601 | 3.79 | 8400 | inf | 0.2519 | 0.0567 |
| 0.1546 | 3.98 | 8800 | inf | 0.2487 | 0.0554 |
| 0.1395 | 4.16 | 9200 | inf | 0.2449 | 0.0551 |
| 0.1364 | 4.34 | 9600 | inf | 0.2425 | 0.0542 |
| 0.1341 | 4.52 | 10000 | inf | 0.2406 | 0.0537 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.2.1
- Datasets 2.17.0
- Tokenizers 0.15.2