Why was st_mod_tsit_factor introduced? #2
Description
Hello,
Thank you for open-sourcing the great work.
Unfortunately, I am not able to understand the reasoning behind "st_mod_tsit_factor"? If I understand correctly, for the small model this parameter is set to 1. And for the two larger models, this parameter was set to 4.
Original Alphafold/Openfold did not have anything like this. As well as, the frame2seq paper does not have any ablations for testing the effect of this parameter.
If possible, it will be really very helpful to understand what thought process went into intoducing this parameter and how did it affect the results.
Thank you,