wav2vec2-xls-r-300m-bangla-command-word-combination-synthetic
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: 0.0068
- Wer: 0.4111
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: 0.0001
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.2982 | 17.86 | 500 | 2.4580 | 1.1089 |
| 0.9644 | 35.71 | 1000 | 0.1250 | 0.5156 |
| 0.1767 | 53.57 | 1500 | 0.0310 | 0.4267 |
| 0.0912 | 71.43 | 2000 | 0.0149 | 0.4178 |
| 0.0505 | 89.29 | 2500 | 0.0068 | 0.4111 |
Framework versions
- Transformers 4.16.2
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.11.0