dgx1_whisper_base_finetune_teacher_no_noise_mozilla_100_epochs_batch_8
This model is a fine-tuned version of openai/whisper-base.en on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.9169
- Wer: 31.4602
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: 0.001
- train_batch_size: 8
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 256
- total_train_batch_size: 2048
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine_with_restarts
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2713 | 14.7 | 500 | 0.7014 | 33.4447 |
| 0.3607 | 29.41 | 1000 | 0.8578 | 34.2037 |
| 0.0171 | 44.12 | 1500 | 0.9436 | 33.6293 |
| 0.0035 | 58.82 | 2000 | 0.9271 | 31.8954 |
| 0.0 | 73.53 | 2500 | 0.9195 | 31.5264 |
| 0.0 | 88.23 | 3000 | 0.9169 | 31.4602 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1
- Datasets 2.8.0
- Tokenizers 0.13.2