wav2vec2-large-xlsr-53-iv_hindi_only-62079-normalized-alldata-1e-4-epochs-100-FT
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6907
- Global Wer: 35.9644
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: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 500
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Global Wer |
|---|---|---|---|---|
| 13.2706 | 1.0 | 540 | 2.5987 | 99.9815 |
| 3.4143 | 2.0 | 1080 | 0.9368 | 51.5660 |
| 2.0940 | 3.0 | 1620 | 0.7728 | 45.2065 |
| 1.7474 | 4.0 | 2160 | 0.7075 | 39.0102 |
| 1.5349 | 5.0 | 2700 | 0.6690 | 37.2794 |
| 1.3923 | 6.0 | 3240 | 0.6491 | 35.6041 |
| 1.2878 | 7.0 | 3780 | 0.6389 | 34.7756 |
| 1.1903 | 8.0 | 4320 | 0.6526 | 34.2983 |
| 1.1363 | 9.0 | 4860 | 0.6532 | 35.7611 |
| 1.1413 | 10.0 | 5400 | 0.6830 | 35.4901 |
| 1.1045 | 11.0 | 5940 | 0.6907 | 35.9644 |
Framework versions
- Transformers 5.13.0
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.22.2