Abhinav Arora

I agree with @parmeet. Dropout can always be used during training regardless of the fact that we freeze the encoder or not.

Thanks @ProGamerGov for reporting this. We are looking to standardize the interface of our transforms/tokenizers and this will be an action item for us to refactor this if needed. The...

> Why is this necessary? > > The current proposed change is touching the code path for nightly build. So I don't think this is correct. > > Has something...