wav2vec2-large-xlsr-turkish-demo-colab
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.4055
- Wer: 0.4800
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 |
---|---|---|---|---|
5.0179 | 4.21 | 400 | 1.4935 | 1.0249 |
0.7075 | 8.42 | 800 | 0.4546 | 0.6071 |
0.3072 | 12.63 | 1200 | 0.3947 | 0.5401 |
0.2145 | 16.84 | 1600 | 0.4049 | 0.5194 |
0.1647 | 21.05 | 2000 | 0.4199 | 0.5003 |
0.1338 | 25.26 | 2400 | 0.4144 | 0.4859 |
0.116 | 29.47 | 2800 | 0.4055 | 0.4800 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.9.1+cu102
- Datasets 1.13.3
- Tokenizers 0.10.3
- Downloads last month
- 25
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.