wav2vec2-hindi-new-4
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 2.3743
- Wer: 0.8926
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: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 40
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 8.0662 | 6.45 | 400 | 3.6328 | 1.0 |
| 2.771 | 12.9 | 800 | 1.5110 | 0.8819 |
| 0.5131 | 19.35 | 1200 | 1.9369 | 0.8989 |
| 0.1955 | 25.8 | 1600 | 2.1664 | 0.8562 |
| 0.1264 | 32.26 | 2000 | 2.3343 | 0.8985 |
| 0.0992 | 38.7 | 2400 | 2.3743 | 0.8926 |
Framework versions
- Transformers 4.20.0.dev0
- Pytorch 1.11.0+cu113
- Datasets 2.2.3.dev0
- Tokenizers 0.12.1