wav2vec2-base-finetune-v2-flickr8k
This model is a fine-tuned version of facebook/wav2vec2-base-960h on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.1698
- Wer: 0.1382
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: 0.0001
- train_batch_size: 12
- eval_batch_size: 2
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.7995 | 1.6 | 1000 | 0.2923 | 0.2001 |
| 0.2981 | 3.2 | 2000 | 0.2247 | 0.1525 |
| 0.2125 | 4.8 | 3000 | 0.1827 | 0.1490 |
| 0.2156 | 6.4 | 4000 | 0.2009 | 0.1445 |
| 0.2783 | 8.0 | 5000 | 0.1678 | 0.1412 |
| 0.1561 | 9.6 | 6000 | 0.1698 | 0.1382 |
Framework versions
- Transformers 4.27.4
- Pytorch 2.0.0
- Datasets 2.11.0
- Tokenizers 0.13.2