facebook

/

esmfold_v1

like 17

Hi,

Why is the special token for padding A in special_tokens_map.json? It seems like the padding token is <pad> in the vocabulary, in the default instantiation of EsmTokenizer, etc.

Is there a reason for this? It seems strange to add random Alanine's to pad a protein so that the whole batch is of the same size.

Thanks in advance for your reply!