wav2vec2-large-xlsr-53-iv_hindi_vaani-62420-hinglish_mixed_scripts-alldata-1e-4-100-mix-0.3-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: 0.7903
- Global Wer: 29.1776
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 |
|---|---|---|---|---|
| 3.7970 | 1.0 | 1080 | 0.9984 | 55.9168 |
| 1.3922 | 2.0 | 2160 | 0.6060 | 38.9238 |
| 0.9264 | 3.0 | 3240 | 0.5426 | 33.9215 |
| 0.6887 | 4.0 | 4320 | 0.5400 | 32.7486 |
| 0.5215 | 5.0 | 5400 | 0.5864 | 31.4466 |
| 0.4202 | 6.0 | 6480 | 0.6417 | 31.1062 |
| 0.3207 | 7.0 | 7560 | 0.6203 | 29.3172 |
| 0.2782 | 8.0 | 8640 | 0.6283 | 29.4289 |
| 0.2306 | 9.0 | 9720 | 0.6740 | 29.5284 |
| 0.2120 | 10.0 | 10800 | 0.7070 | 28.9193 |
| 0.1816 | 11.0 | 11880 | 0.6883 | 28.8320 |
| 0.1638 | 12.0 | 12960 | 0.7168 | 28.8128 |
| 0.1410 | 13.0 | 14040 | 0.7246 | 29.3207 |
| 0.1436 | 14.0 | 15120 | 0.7130 | 29.4673 |
| 0.1211 | 15.0 | 16200 | 0.7564 | 28.7796 |
| 0.1181 | 16.0 | 17280 | 0.7765 | 28.5283 |
| 0.1068 | 17.0 | 18360 | 0.7739 | 29.1863 |
| 0.1068 | 18.0 | 19440 | 0.8337 | 29.0519 |
| 0.0971 | 19.0 | 20520 | 0.7903 | 29.1776 |
Framework versions
- Transformers 5.13.0
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.22.2