when the protein length is larger than 1022, the recommend output will be staggered by two digits

[liusfore]

here is the output, and the number of residues output by the language model is reduced by two.

F1037Y F1038S Y1039N S1040I N1041M I1042N M1043F N1044F F1045K F1046T K1047E T1048I E1049T I1050L T1051A L1052N A1053G N1054E G1055I E1056R I1057K K1059P R1060L P1061I L1062E I1063T E1064N T1065G N1066E G1067T E1068G T1069E G1070I E1071V V1073D W1074K D1075G

[brianhie]

This may be a bug with the way we are handling sequences longer than the ESM context length, will investigate.