metadata
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- common_voice
model-index:
- name: wav2vec2-xls-r-300m-ar
results: []
wav2vec2-xls-r-300m-ar
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 21.9388
- Wer: 1.0
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: 7.5e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 18.7839 | 0.5 | 1 | 32.5547 | 1.0 |
| 19.4045 | 1.0 | 2 | 32.5547 | 1.0 |
| 19.6279 | 1.5 | 3 | 31.8850 | 1.0 |
| 17.7216 | 2.0 | 4 | 31.1082 | 1.0 |
| 17.3103 | 2.5 | 5 | 31.1082 | 1.0 |
| 17.2026 | 3.0 | 6 | 29.9666 | 1.0 |
| 16.7911 | 3.5 | 7 | 28.5444 | 1.0 |
| 13.7364 | 4.0 | 8 | 25.7932 | 1.0 |
| 13.1602 | 4.5 | 9 | 23.4634 | 1.0 |
| 12.9439 | 5.0 | 10 | 21.9388 | 1.0 |
Framework versions
- Transformers 4.17.0.dev0
- Pytorch 1.10.2+cu102
- Datasets 1.18.2.dev0
- Tokenizers 0.11.0