Whisper Large-v2 Czech CV11 audio concatenation test
This model is a fine-tuned version of openai/whisper-large-v2 on the mozilla-foundation/common_voice_11_0 cs dataset. It achieves the following results on the evaluation set:
- Loss: 0.2563
- Wer: 8.3774
Model description
First test of audio concatenation few short samples to one training sample together.
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: 32
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 2
- 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
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0022 | 24.39 | 1000 | 0.2181 | 8.7807 |
| 0.0002 | 48.77 | 2000 | 0.2563 | 8.3774 |
| 0.0001 | 73.17 | 3000 | 0.2756 | 8.4510 |
| 0.0001 | 97.55 | 4000 | 0.2871 | 8.4823 |
| 0.0001 | 121.94 | 5000 | 0.2913 | 8.4731 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.0+cu117
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2