fine-tune-wav2vec2-large-swahili
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.9579
- Wer: 0.5860
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- 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
- num_epochs: 9
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
No log | 1.16 | 200 | 2.9678 | 1.0 |
4.5893 | 2.33 | 400 | 2.9616 | 1.0 |
4.5893 | 3.49 | 600 | 2.9572 | 1.0 |
2.9305 | 4.65 | 800 | 2.8201 | 1.0 |
2.9305 | 5.81 | 1000 | 1.3537 | 0.8343 |
1.7413 | 6.98 | 1200 | 1.0854 | 0.6862 |
1.7413 | 8.14 | 1400 | 0.9579 | 0.5860 |
Framework versions
- Transformers 4.27.0
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
- Downloads last month
- 3