Wave2Vec2-Bert2.0 - Kiran Pantha
This model is a fine-tuned version of kiranpantha/w2v-bert-2.0-nepali on the kiranpantha/OpenSLR54-Balanced-Nepali dataset. It achieves the following results on the evaluation set:
- Loss: 0.3739
- Wer: 0.4058
- Cer: 0.0951
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.643 | 0.24 | 300 | 0.4271 | 0.4079 | 0.0919 |
| 0.6342 | 0.48 | 600 | 0.4928 | 0.4902 | 0.1245 |
| 0.6421 | 0.72 | 900 | 0.4251 | 0.4595 | 0.1112 |
| 0.5773 | 0.96 | 1200 | 0.4170 | 0.4342 | 0.1069 |
| 0.5107 | 1.2 | 1500 | 0.4487 | 0.4469 | 0.1089 |
| 0.4639 | 1.44 | 1800 | 0.3823 | 0.4157 | 0.0973 |
| 0.4369 | 1.6800 | 2100 | 0.3792 | 0.4145 | 0.0984 |
| 0.449 | 1.92 | 2400 | 0.3739 | 0.4058 | 0.0951 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.5.0+cu124
- Datasets 3.0.2
- Tokenizers 0.20.1