Added F and H functions in addition to augmented lagrangian to enable automated learning provided the client passes valid pytorch expressions f_tch and [g_tch] along with objective and constraints. Code still needs a lot of work (for example, we can only process a single uncertainty parameter u as of now, and there is a ton of redundancy), but this will enable automated learning in a much cleaner way on the client side for the time being.
Added F and H functions in addition to augmented lagrangian to enable automated learning provided the client passes valid pytorch expressions f_tch and [g_tch] along with objective and constraints. Code still needs a lot of work (for example, we can only process a single uncertainty parameter u as of now, and there is a ton of redundancy), but this will enable automated learning in a much cleaner way on the client side for the time being.