<!-- 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. -->
wav2vec2-xlsr-ft-cv13-cy
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the MOZILLA-FOUNDATION/COMMON_VOICE_13_0 - CY dataset. It achieves the following results on the evaluation set:
- Loss: 0.4552
- Wer: 0.2529
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: 32
- 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: 800
- num_epochs: 30.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.2113 | 1.95 | 400 | 0.8211 | 0.7142 |
| 0.5062 | 3.9 | 800 | 0.4197 | 0.4302 |
| 0.2767 | 5.85 | 1200 | 0.3882 | 0.3764 |
| 0.1793 | 7.8 | 1600 | 0.3928 | 0.3453 |
| 0.132 | 9.76 | 2000 | 0.4073 | 0.3331 |
| 0.1071 | 11.71 | 2400 | 0.4364 | 0.3138 |
| 0.0933 | 13.66 | 2800 | 0.4310 | 0.2973 |
| 0.0816 | 15.61 | 3200 | 0.4229 | 0.2926 |
| 0.0735 | 17.56 | 3600 | 0.4422 | 0.2889 |
| 0.0665 | 19.51 | 4000 | 0.4509 | 0.2821 |
| 0.0583 | 21.46 | 4400 | 0.4678 | 0.2734 |
| 0.0529 | 23.41 | 4800 | 0.4459 | 0.2655 |
| 0.048 | 25.37 | 5200 | 0.4466 | 0.2620 |
| 0.0428 | 27.32 | 5600 | 0.4565 | 0.2579 |
| 0.0422 | 29.27 | 6000 | 0.4550 | 0.2532 |
Framework versions
- Transformers 4.29.2
- Pytorch 2.0.1+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3
