Usability questions

[dMedinaO]

Dear community,

Thanks for your work. I am very interesting to use tape and pretrained models for landscape navigation to discovery new protein sequences.

I have achieved to run the bet pretrained model, nevertheless, I would like to explore other pretrained models like unirep, trrosseta, or transformer, but i have not been successful, can you give me some example of how to use these pretrained models?

Other questions, once I have the encoding sequences, obviously, I would to like my ml model to predict new functions or properties depending on the problem, once the models is completed, I would to explore new sequences, normally, I can simulate mutations and then apply the numerical representation strategy, and then the predictive model. But, is there some strategy, implemented function in tape, or similar, to decode numerical representations (the output to the pretrained model) for obtaining a protein sequence?

Thanks so much.