indicwav2vec-hindi-mucs-61961-hinglish_mixed_scripts-alldata-1e-4-steps-12000-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: 1.1530
- Global Wer: 35.3555
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: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- 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 |
|---|---|---|---|---|
| 2.5364 | 0.1594 | 1000 | 1.4022 | 49.1797 |
| 2.0891 | 0.3188 | 2000 | 1.2507 | 44.0041 |
| 2.0087 | 0.4782 | 3000 | 1.2103 | 44.0943 |
| 1.8096 | 0.6376 | 4000 | 1.2069 | 39.1470 |
| 1.8881 | 0.7969 | 5000 | 1.1543 | 38.4310 |
| 1.6947 | 0.9563 | 6000 | 1.1616 | 38.1519 |
| 1.6307 | 1.1157 | 7000 | 1.1421 | 36.5423 |
| 1.6378 | 1.2751 | 8000 | 1.1412 | 37.7995 |
| 1.6743 | 1.4345 | 9000 | 1.1660 | 35.9926 |
| 1.5278 | 1.5939 | 10000 | 1.1973 | 35.4006 |
| 1.4878 | 1.7533 | 11000 | 1.1275 | 35.8516 |
| 1.5591 | 1.9127 | 12000 | 1.1530 | 35.3555 |
Framework versions
- Transformers 5.13.0
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.22.2