<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wa2vec2-large-xls-r-colab_turkish
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: 0.3941
- Wer: 0.3812
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: 30
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
4.0265 | 3.67 | 400 | 0.7368 | 0.8192 |
0.4253 | 7.34 | 800 | 0.4467 | 0.5111 |
0.1902 | 11.01 | 1200 | 0.4423 | 0.4723 |
0.1293 | 14.68 | 1600 | 0.3854 | 0.4216 |
0.0989 | 18.35 | 2000 | 0.3997 | 0.4197 |
0.0745 | 22.02 | 2400 | 0.4133 | 0.4182 |
0.0598 | 25.69 | 2800 | 0.3962 | 0.3925 |
0.0488 | 29.36 | 3200 | 0.3941 | 0.3812 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.7.1
- Tokenizers 0.13.2