Inference multiple protein structure with single compilation

As you stated in the project README (in Inferencing many proteins), I am inferencing many small multimer structures.
right now I precompute MSA , and predict each poly-peptide in separate python execution.
In small protein I can see in the bench mark mode that sometimes 2/3 of the inferencing time is during compilation.

Do you have any open issue or intent to add the padding option when tring to predict more than one protein, with the same compilation?