subhadeep_whisper_medium_finetune_teacher_babble_noise_mozilla_original_case_100_epochs_batch_2
This model is a fine-tuned version of openai/whisper-medium.en on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5402
- Wer: 33.4656
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 128
- total_train_batch_size: 512
- 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 |
|---|---|---|---|---|
| 1.5492 | 0.73 | 100 | 0.4618 | 24.3238 |
| 0.2737 | 1.47 | 200 | 0.4035 | 25.2846 |
| 0.2041 | 2.2 | 300 | 0.4131 | 24.3864 |
| 0.1025 | 2.94 | 400 | 0.4408 | 25.2289 |
| 0.0592 | 3.67 | 500 | 0.4820 | 25.9948 |
| 0.0577 | 4.41 | 600 | 0.5402 | 33.4656 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1
- Datasets 2.8.0
- Tokenizers 0.13.2