indicwav2vec-hindi-mucs-62256-hinglish_mixed_scripts-25_45-1e-4-epochs-100-FT
This model is a fine-tuned version of ai4bharat/indicwav2vec-hindi on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.9322
- Global Wer: 22.1205
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 |
|---|---|---|---|---|
| No log | 1.0 | 440 | 1.7133 | 65.0212 |
| 18.0510 | 2.0 | 880 | 1.0150 | 37.6376 |
| 1.7629 | 3.0 | 1320 | 0.9221 | 31.1649 |
| 1.3020 | 4.0 | 1760 | 0.8467 | 29.5739 |
| 1.1410 | 5.0 | 2200 | 0.8176 | 30.9360 |
| 0.9985 | 6.0 | 2640 | 0.8306 | 27.1222 |
| 0.9000 | 7.0 | 3080 | 0.7844 | 27.1657 |
| 0.8144 | 8.0 | 3520 | 0.7761 | 25.3896 |
| 0.8144 | 9.0 | 3960 | 0.7721 | 25.6620 |
| 0.7349 | 10.0 | 4400 | 0.7903 | 24.7140 |
| 0.6751 | 11.0 | 4840 | 0.7840 | 24.5723 |
| 0.6202 | 12.0 | 5280 | 0.7774 | 24.2672 |
| 0.5885 | 13.0 | 5720 | 0.8225 | 23.3083 |
| 0.5256 | 14.0 | 6160 | 0.8127 | 23.6788 |
| 0.5215 | 15.0 | 6600 | 0.8227 | 22.8615 |
| 0.4809 | 16.0 | 7040 | 0.8543 | 22.7525 |
| 0.4809 | 17.0 | 7480 | 0.8370 | 22.4583 |
| 0.4547 | 18.0 | 7920 | 0.9087 | 21.8917 |
| 0.4305 | 19.0 | 8360 | 0.9018 | 22.5455 |
| 0.4043 | 20.0 | 8800 | 0.8708 | 22.5346 |
| 0.4003 | 21.0 | 9240 | 0.9322 | 22.1205 |
Framework versions
- Transformers 5.13.0
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.22.2