wav2vec2-large-xls-r-300m-kika_my-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 1.3300
- Wer: 0.5804
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
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.8067 | 4.82 | 400 | 1.2892 | 0.8886 |
| 0.3048 | 9.64 | 800 | 1.2285 | 0.6797 |
| 0.1413 | 14.46 | 1200 | 1.1970 | 0.6509 |
| 0.1047 | 19.28 | 1600 | 1.3628 | 0.6166 |
| 0.0799 | 24.1 | 2000 | 1.3345 | 0.6014 |
| 0.0638 | 28.92 | 2400 | 1.3300 | 0.5804 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.17.0
- Tokenizers 0.10.3