You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The NonlinearConstraint class currently only supports constraint functions that return a single output. The user can still provide a list of constraints in the search space, but each nonlinear instance can contain only a single constraint. Even though the trieste code itself supports multi-output constraints, the tfp.math.value_and_gradient call used in NonlinearConstraint.__init__ does not work correctly. It returns the sum of gradients instead of the Jacobian needed. It will likely require a custom gradient implementation to support this correctly.
The text was updated successfully, but these errors were encountered:
The
NonlinearConstraint
class currently only supports constraint functions that return a single output. The user can still provide a list of constraints in the search space, but each nonlinear instance can contain only a single constraint. Even though the trieste code itself supports multi-output constraints, thetfp.math.value_and_gradient
call used inNonlinearConstraint.__init__
does not work correctly. It returns the sum of gradients instead of the Jacobian needed. It will likely require a custom gradient implementation to support this correctly.The text was updated successfully, but these errors were encountered: