metadata
license: apache-2.0
base_model: facebook/wav2vec2-large-xlsr-53
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: XLS-R_Jibbali_lang
results: []
XLS-R_Jibbali_lang
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1714
- Wer: 0.1941
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: 0.0003
- 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: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 16.1473 | 0.99 | 56 | 11.0651 | 1.0 |
| 3.9177 | 2.0 | 113 | 3.4720 | 1.0 |
| 3.1902 | 2.99 | 169 | 3.1574 | 1.0 |
| 3.1711 | 4.0 | 226 | 3.1379 | 1.0 |
| 3.1491 | 4.99 | 282 | 3.1154 | 1.0 |
| 3.1449 | 6.0 | 339 | 3.0533 | 1.0 |
| 2.9214 | 6.99 | 395 | 2.7533 | 1.0 |
| 1.9003 | 8.0 | 452 | 1.4168 | 0.9291 |
| 0.6151 | 8.99 | 508 | 0.3110 | 0.3224 |
| 0.2125 | 10.0 | 565 | 0.2170 | 0.2145 |
| 0.1754 | 10.99 | 621 | 0.1987 | 0.2069 |
| 0.1688 | 12.0 | 678 | 0.1870 | 0.1985 |
| 0.1012 | 12.99 | 734 | 0.1856 | 0.1908 |
| 0.1157 | 14.0 | 791 | 0.1906 | 0.2025 |
| 0.1427 | 14.99 | 847 | 0.1844 | 0.1937 |
| 0.0513 | 16.0 | 904 | 0.1852 | 0.1915 |
| 0.1403 | 16.99 | 960 | 0.1713 | 0.1944 |
| 0.1119 | 18.0 | 1017 | 0.1610 | 0.1974 |
| 0.1034 | 18.99 | 1073 | 0.1697 | 0.1944 |
| 0.0428 | 19.82 | 1120 | 0.1714 | 0.1941 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu121
- Datasets 2.17.0
- Tokenizers 0.15.2