wav2vec2-large-xls-r-300m-irish-colab
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: 1.4286
- Wer: 0.5097
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: 32
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 210
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 4.3406 | 24.97 | 400 | 1.1677 | 0.7270 |
| 0.2527 | 49.97 | 800 | 1.2686 | 0.5927 |
| 0.0797 | 74.97 | 1200 | 1.3970 | 0.5769 |
| 0.0424 | 99.97 | 1600 | 1.4093 | 0.5600 |
| 0.0286 | 124.97 | 2000 | 1.3684 | 0.5407 |
| 0.0174 | 149.97 | 2400 | 1.4571 | 0.5205 |
| 0.0109 | 174.97 | 2800 | 1.4327 | 0.5178 |
| 0.0072 | 199.97 | 3200 | 1.4286 | 0.5097 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu113
- Datasets 1.13.3
- Tokenizers 0.10.3