wav2vec2-large-xlsr-53-indic_voices-61925-normalized-30_70-1e-4-steps-12000-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: 0.7691
- Global Wer: 53.1980
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: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 2
- 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
- training_steps: 12000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Global Wer |
|---|---|---|---|---|
| 6.9159 | 0.0868 | 600 | 3.3896 | 100.0 |
| 5.2926 | 0.1737 | 1200 | 2.2943 | 99.8717 |
| 3.6771 | 0.2605 | 1800 | 1.5508 | 84.7990 |
| 3.2207 | 0.3473 | 2400 | 1.2201 | 76.0665 |
| 2.7807 | 0.4342 | 3000 | 1.0720 | 66.4195 |
| 2.5455 | 0.5210 | 3600 | 1.0258 | 65.8730 |
| 2.4389 | 0.6079 | 4200 | 0.9466 | 61.2948 |
| 2.3633 | 0.6947 | 4800 | 0.9265 | 60.7316 |
| 2.4801 | 0.7815 | 5400 | 0.8723 | 58.6740 |
| 2.3967 | 0.8684 | 6000 | 0.8700 | 57.6033 |
| 2.3953 | 0.9552 | 6600 | 0.8389 | 55.7910 |
| 1.9146 | 1.0420 | 7200 | 0.8140 | 54.5363 |
| 1.9893 | 1.1288 | 7800 | 0.8353 | 56.3765 |
| 1.9715 | 1.2156 | 8400 | 0.7902 | 53.9452 |
| 2.0774 | 1.3025 | 9000 | 0.7763 | 52.2668 |
| 2.2356 | 1.3893 | 9600 | 0.7976 | 52.8244 |
| 2.0103 | 1.4762 | 10200 | 0.8067 | 53.4601 |
| 1.9861 | 1.5630 | 10800 | 0.7691 | 53.1980 |
Framework versions
- Transformers 5.13.0
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.22.2