translation of the model for use in faster-whisper.
Hello! I think you are familiar with faster-whisper algorithms. The developer of this software has his own site on GitHub, where he uses the same large-v2 and other models, only converts them first, and even attached software in Python, with which you can convert ordinary whisper ones.As a Russian user, I really liked your work, your model! but I can’t translate it for use in faster-whisper (for my work in law, it is preferable to work with this version). As the developer of your Russian model, can you try to translate your model for faster-whisper? for you as a developer, it may be an interesting experience and you will help me out a lot! Thank you. here is his github page
https://github.com/guillaumekln/faster-whisper#model-conversion
I myself transformed your model for use in faster whisper, but after that it worked poorly.
I'm sorry, I'm not really familiar with faster-whisper, so I'm not sure how to do the conversion process.
I would suggest simply using the v3 model once that is converted (which I assume it likely is), as that will likely have better Russian performance.