Whisper Medium nl - GeoffVdr
This model is a fine-tuned version of openai/whisper-medium on the Common Voice 11.0 dataset.
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
- Training: Mozilla CommonVoice 11 Dutch train+validation set
- Evaluation: Mozilla CommonVoice 11 Dutch test set
Training procedure
Training Hyperparameters
- learning_rate: 1e-5
- train_batch_size: 8
- eval_batch_size: 8
- gradient_accumulation_steps: 2
- lr_scheduler_warmup_steps: 500
- training_steps: 12000
Training Results
| Training Loss | Epoch | Step | WER |
|---|---|---|---|
| 0.1111 | 0.39 | 1000 | 9.89 |
| 0.0884 | 0.78 | 2000 | 9.26 |
| 0.0362 | 1.17 | 3000 | 8.64 |
| 0.0359 | 1.56 | 4000 | 8.60 |
| 0.0375 | 1.95 | 5000 | 8.24 |
| : | |||
| : | |||
| 0.0015 | 4.68 | 12000 | 7.51 |
Framework versions
- Downloads last month
- 5
This model does not have enough activity to be deployed to Inference API (serverless) yet.
Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Dataset used to train GeoffVdr/whisper-medium-nlcv11
Space using GeoffVdr/whisper-medium-nlcv11 1
Evaluation results
- Wer on mozilla-foundation/common_voice_11_0test set self-reported7.514