Whisper Large v2 PL
This model is a fine-tuned version of openai/whisper-large-v2 on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.4222
- Wer: 6.9125
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: 1e-05
- train_batch_size: 8
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 8
- 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
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1144 | 1.93 | 500 | 0.2016 | 7.4749 |
| 0.0441 | 3.86 | 1000 | 0.2193 | 7.3154 |
| 0.0099 | 5.79 | 1500 | 0.2983 | 7.0804 |
| 0.0048 | 7.72 | 2000 | 0.3514 | 7.0988 |
| 0.0017 | 9.65 | 2500 | 0.3614 | 7.0485 |
| 0.0014 | 11.58 | 3000 | 0.3814 | 7.1240 |
| 0.001 | 13.51 | 3500 | 0.3773 | 6.9931 |
| 0.0005 | 15.44 | 4000 | 0.4085 | 6.9662 |
| 0.0004 | 17.37 | 4500 | 0.4195 | 6.9192 |
| 0.0004 | 19.3 | 5000 | 0.4222 | 6.9125 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.0+cu117
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2