search

lucidrains / audiolm-pytorch

Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch
MIT License
2.36k stars 255 forks source link

changed to s3 support #138

Closed marianna13 closed 1 year ago

marianna13 commented 1 year ago

Broken dataloader.

lucidrains commented 1 year ago

@marianna13 oh i see, thanks for this Marianna! i will do an in-depth review early next week and we can try to get it merged next friday? looking at the changes, it may need some work before i can accept

marianna13 commented 1 year ago

@lucidrains no don't merge it yet, it's broken, I just want to make it work

lucidrains commented 1 year ago

@marianna13 ok cool! support for Romain's cc2dataset would definitely be ideal at some point

you can mark your pull request as a draft, if you want the maintainer to know it isn't ready for review. pro tip

lucidrains commented 1 year ago

@marianna13 while you try to get it functional, i can come meet you from the other end

maybe for starters allow for custom train and valid datasets to be passed into the trainers?

marianna13 commented 1 year ago

@lucidrains yeah it would be nice! thanks