wav2vec2-large-xls-r-300m-turkish-colab-full
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: 0.3991
- Wer: 0.3050
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
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.9196 | 3.67 | 400 | 0.6539 | 0.6524 |
| 0.3908 | 7.34 | 800 | 0.4486 | 0.4502 |
| 0.1859 | 11.01 | 1200 | 0.4015 | 0.3799 |
| 0.1228 | 14.68 | 1600 | 0.4080 | 0.3741 |
| 0.0956 | 18.35 | 2000 | 0.3930 | 0.3468 |
| 0.0757 | 22.02 | 2400 | 0.4163 | 0.3355 |
| 0.0573 | 25.69 | 2800 | 0.3983 | 0.3115 |
| 0.0463 | 29.36 | 3200 | 0.3991 | 0.3050 |
Framework versions
- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 1.18.3
- Tokenizers 0.13.3