<!-- 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-en-cy-btb
This model is a fine-tuned version of techiaith/wav2vec2-xlsr-ft-en-cy on the TECHIAITH/BANC-TRAWSGRIFIADAU-BANGOR - NA dataset. It achieves the following results on the evaluation set:
- Loss: 0.3819
- Wer: 0.2847
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: 5.0
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
No log | 0.21 | 100 | 3.0597 | 1.0 |
No log | 0.41 | 200 | 2.0286 | 0.9948 |
No log | 0.62 | 300 | 0.7111 | 0.4665 |
No log | 0.83 | 400 | 0.5970 | 0.4187 |
2.4806 | 1.03 | 500 | 0.5456 | 0.3909 |
2.4806 | 1.24 | 600 | 0.5525 | 0.3886 |
2.4806 | 1.44 | 700 | 0.4990 | 0.3861 |
2.4806 | 1.65 | 800 | 0.4893 | 0.3723 |
2.4806 | 1.86 | 900 | 0.4667 | 0.3538 |
0.6124 | 2.06 | 1000 | 0.4443 | 0.3371 |
0.6124 | 2.27 | 1100 | 0.4412 | 0.3377 |
0.6124 | 2.48 | 1200 | 0.4290 | 0.3281 |
0.6124 | 2.68 | 1300 | 0.4185 | 0.3213 |
0.6124 | 2.89 | 1400 | 0.4106 | 0.3199 |
0.4761 | 3.1 | 1500 | 0.4151 | 0.3177 |
0.4761 | 3.3 | 1600 | 0.4085 | 0.3092 |
0.4761 | 3.51 | 1700 | 0.4156 | 0.3109 |
0.4761 | 3.72 | 1800 | 0.3956 | 0.3020 |
0.4761 | 3.92 | 1900 | 0.3899 | 0.2984 |
0.3826 | 4.13 | 2000 | 0.3931 | 0.2953 |
0.3826 | 4.33 | 2100 | 0.3898 | 0.2892 |
0.3826 | 4.54 | 2200 | 0.3828 | 0.2862 |
0.3826 | 4.75 | 2300 | 0.3866 | 0.2876 |
0.3826 | 4.95 | 2400 | 0.3817 | 0.2854 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.3