Request for checkpoint files
Hey, I have been trying to train my own JARVIS voice, but your work is far better!
Will you be able to share the checkpoint files? I'd like you fine tune these even further.
Thanks!
@praharshbhatt I'm traveling this week, but when I get back I can upload the checkpoints to this repo. If I forget, feel free to ping me again here as a reminder :)
@jgkawell
thanks a ton for getting back to me!
May I also ask what's the training data that you've used?
I'm using these wavs, but my models have distorted voices even at 1200 epochs.
@praharshbhatt
Just uploaded the checkpoints to the repo. Check them out and let me know if they work for you. I mentioned the training data in #1. Your dataset looks similar to mine and I just uploaded my metadata.csv which has my transcription of all the training data so hope that helps. I'm not sharing the raw audio for what I'm sure are obvious reasons. Keep in mind that the high model is based off of lessac and the medium model is based off of alan and I don't think you'll ever get good results trying to train from scratch on such a limited dataset (check MODEL_CARDs for details).