This PR fixes a bug in the relative_step function where the step was computed by averaging over one axis of a parameter. When the parameter is a morphology, this caused vertical artifacts to appear during the optimisation process.
I also used this PR to remove a few variables that were not used.
This PR fixes a bug in the relative_step function where the step was computed by averaging over one axis of a parameter. When the parameter is a morphology, this caused vertical artifacts to appear during the optimisation process.
I also used this PR to remove a few variables that were not used.