Whisper Medium Serbian - Drishti Sharma
This model is a fine-tuned version of openai/whisper-medium on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3873
- Wer: 11.6147
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: 32
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 800
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.1691 | 1.89 | 100 | 0.2398 | 13.5977 |
0.0571 | 3.77 | 200 | 0.2419 | 12.7479 |
0.0225 | 5.66 | 300 | 0.2869 | 12.2218 |
0.0108 | 7.55 | 400 | 0.3085 | 12.2622 |
0.0121 | 9.43 | 500 | 0.3478 | 11.6147 |
0.0015 | 11.32 | 600 | 0.3702 | 11.5338 |
0.0005 | 13.21 | 700 | 0.3839 | 11.6147 |
0.0005 | 15.09 | 800 | 0.3873 | 11.6147 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.0+cu116
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2
- Downloads last month
- 2
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.