[todo] structure module

[lucidrains]

in the structure module, the FAPE loss needs to be applied on every iteration, while the rotation needs to have a stop-gradient for all but the last iteration

[hypnopump]

May i ask what "the rotation needs to have a stop-gradient for all but the last iteration" means?

The FAPE function satisfies the main constraints the methods section outlined: https://github.com/EleutherAI/mp_nerf/blob/master/mp_nerf/ml_utils.py#L102 Do you find something to be missing? i might add it

[lucidrains]

May i ask what "the rotation needs to have a stop-gradient for all but the last iteration" means?

The FAPE function satisfies the main constraints the methods section outlined: https://github.com/EleutherAI/mp_nerf/blob/master/mp_nerf/ml_utils.py#L102 Do you find something to be missing? i might add it

5214156 took care of it here :)

just need to apply the FAPE loss every iteration and sum up the auxiliary losses now!

[lucidrains]

May i ask what "the rotation needs to have a stop-gradient for all but the last iteration" means?

The FAPE function satisfies the main constraints the methods section outlined: https://github.com/EleutherAI/mp_nerf/blob/master/mp_nerf/ml_utils.py#L102 Do you find something to be missing? i might add it

they have this MultiRigidSidechain class https://github.com/deepmind/alphafold/blob/0bab1bf84d9d887aba5cfb6d09af1e8c3ecbc408/alphafold/model/folding.py#L931 and it seems like there's a parameterized transformation before applying the loss itself