metadata
license: mit
base_model: facebook/w2v-bert-2.0
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v-bert-Marathi-large
results: []
w2v-bert-Marathi-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: 0.190338
- Wer: 0.108757
- Cer: 0.024650
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 |
|---|---|---|---|---|---|
| 2.8076 | 0.5882 | 300 | 0.5988 | 0.5285 | 0.1285 |
| 0.4551 | 1.1765 | 600 | 0.4358 | 0.3706 | 0.0871 |
| 0.3345 | 1.7647 | 900 | 0.3568 | 0.3610 | 0.0779 |
| 0.2521 | 2.3529 | 1200 | 0.3093 | 0.2636 | 0.0581 |
| 0.1886 | 2.9412 | 1500 | 0.2731 | 0.2421 | 0.0541 |
| 0.1352 | 3.5294 | 1800 | 0.2458 | 0.1907 | 0.0419 |
| 0.0951 | 4.1176 | 2100 | 0.2165 | 0.1712 | 0.0363 |
| 0.0608 | 4.7059 | 2400 | 0.2203 | 0.1356 | 0.0303 |
| 0.0348 | 5.2941 | 2700 | 0.2000 | 0.1169 | 0.0260 |
| 0.0166 | 5.8824 | 3000 | 0.1903 | 0.1088 | 0.0247 |
Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.19.2
- Tokenizers 0.19.1