metadata
license: mit
base_model: facebook/w2v-bert-2.0
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v-bert-Tamil-large
results: []
w2v-bert-Tamil-large
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: 0.1815
- Wer: 0.2176
- Cer: 0.0328
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: 2
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 3000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 2.0779 | 0.75 | 300 | 0.4934 | 0.6338 | 0.1189 |
| 0.3653 | 1.5 | 600 | 0.4045 | 0.5424 | 0.0975 |
| 0.2632 | 2.25 | 900 | 0.3148 | 0.4421 | 0.0723 |
| 0.2084 | 3.0 | 1200 | 0.2297 | 0.3499 | 0.0576 |
| 0.1359 | 3.75 | 1500 | 0.2042 | 0.3060 | 0.0464 |
| 0.1049 | 4.5 | 1800 | 0.1939 | 0.2836 | 0.0446 |
| 0.0823 | 5.25 | 2100 | 0.1827 | 0.2504 | 0.0382 |
| 0.0561 | 6.0 | 2400 | 0.1731 | 0.2419 | 0.0368 |
| 0.0352 | 6.75 | 2700 | 0.1802 | 0.2275 | 0.0335 |
| 0.0224 | 7.5 | 3000 | 0.1815 | 0.2176 | 0.0328 |
Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1