wav2vec2-large-xls-r-300m-urdu-cv8-200epochs
This model was trained from scratch on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 1.3200
- Wer: 0.7723
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: 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: 100
- num_epochs: 13
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.3204 | 1.27 | 32 | 1.3200 | 0.7723 |
| 0.3021 | 2.55 | 64 | 1.3200 | 0.7723 |
| 0.3153 | 3.82 | 96 | 1.3200 | 0.7723 |
| 0.3239 | 5.12 | 128 | 1.3200 | 0.7723 |
| 0.3153 | 6.39 | 160 | 1.3200 | 0.7723 |
| 0.3202 | 7.67 | 192 | 1.3200 | 0.7723 |
| 0.3126 | 8.94 | 224 | 1.3200 | 0.7723 |
| 0.3183 | 10.24 | 256 | 1.3200 | 0.7723 |
| 0.3135 | 11.51 | 288 | 1.3200 | 0.7723 |
| 0.3137 | 12.78 | 320 | 1.3200 | 0.7723 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.12.1