Inputs types while pretraining alone about protein_bert HOT 3 CLOSED

Just to make sure I understand the context of your question, you are essentially writing your own code for pretraining the model on your own dataset, right? If so, I think this short thread will be useful for you.

Overall these inputs seem reasonable (assuming you don't have GO annotations), but that would of course depend on how exactly you use them to pretrain the model.

from protein_bert.

@nadavbra thank you for your answer.
Yes I am changing your code a bit. I saw this thread and I also changed the DatasetHandler object to this one -

BTW, what exactly did you mean by "how exactly you use them to pretrain your model?", are you talking about the GO annotations? I dont want to use any at all

Thanks

from protein_bert.

I just meant the overall flow of your code. Everything you do seems reasonable, but it just depends on your overall implementation.
I'd be curious to know whether it ended up working in the end.

from protein_bert.