Fine tuned Whisper ASR by using Open CV Hindi Dataset
- Transformers 4.26.0.dev0
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
Deployment of the above setup can be found here at Huggingface: https://huggingface.co/spaces/Ranjit/Ranjit-whisper-small-hi
The model prepared here is a fine-tuned version of openai/whisper-small model on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:
Loss: 0.4292
Wer: 32.0452
More details about the model can be found here.