wav2vec2bert-bengali-0db-noisy-audio-1e5
This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.8461
- Wer: 0.5885
- Cer: 0.1966
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: 1e-05
- train_batch_size: 32
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 14.1057 | 1.2675 | 100 | 6.9213 | 1.0 | 0.9949 |
| 6.2751 | 2.5350 | 200 | 3.0367 | 0.8988 | 0.3379 |
| 4.4437 | 3.8025 | 300 | 2.2319 | 0.7150 | 0.2458 |
| 3.2883 | 5.0637 | 400 | 1.9939 | 0.6459 | 0.2175 |
| 3.0549 | 6.3312 | 500 | 1.8929 | 0.6151 | 0.2076 |
| 2.3529 | 7.5987 | 600 | 1.8636 | 0.5959 | 0.2024 |
| 2.4048 | 8.8662 | 700 | 1.8461 | 0.5885 | 0.1966 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2