metadata
license: mit
base_model: facebook/w2v-bert-2.0
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: w2v-bert-2.0-nonstudio_and_studioRecords
results: []
w2v-bert-2.0-nonstudio_and_studioRecords
This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1641
- Wer: 0.1184
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: 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: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.1077 | 0.46 | 600 | 0.4029 | 0.4897 |
| 0.1727 | 0.92 | 1200 | 0.2339 | 0.3573 |
| 0.1224 | 1.38 | 1800 | 0.2159 | 0.3225 |
| 0.1103 | 1.84 | 2400 | 0.1838 | 0.2764 |
| 0.0907 | 2.3 | 3000 | 0.1844 | 0.2603 |
| 0.0796 | 2.76 | 3600 | 0.1829 | 0.2498 |
| 0.0685 | 3.22 | 4200 | 0.1719 | 0.2336 |
| 0.0588 | 3.68 | 4800 | 0.1607 | 0.2030 |
| 0.054 | 4.14 | 5400 | 0.1611 | 0.1941 |
| 0.0424 | 4.6 | 6000 | 0.1536 | 0.1821 |
| 0.0402 | 5.06 | 6600 | 0.1562 | 0.1769 |
| 0.0312 | 5.52 | 7200 | 0.1494 | 0.1655 |
| 0.0303 | 5.98 | 7800 | 0.1471 | 0.1510 |
| 0.0218 | 6.44 | 8400 | 0.1707 | 0.1488 |
| 0.0218 | 6.9 | 9000 | 0.1458 | 0.1296 |
| 0.0151 | 7.36 | 9600 | 0.1424 | 0.1326 |
| 0.014 | 7.82 | 10200 | 0.1406 | 0.1266 |
| 0.0107 | 8.28 | 10800 | 0.1476 | 0.1291 |
| 0.0078 | 8.74 | 11400 | 0.1563 | 0.1254 |
| 0.007 | 9.2 | 12000 | 0.1528 | 0.1197 |
| 0.0041 | 9.66 | 12600 | 0.1641 | 0.1184 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.1.1+cu121
- Datasets 2.16.1
- Tokenizers 0.15.1