In this release, - we update the functionals by using the latest autofd. - we make the API simpler, no need to pass `polarized=True/False` anymore, it will be determined from whether `rho` returns shape `()` or `(2,)`
0.0.8
This release adds experimental support for functional derivative, currently awaiting autofd to release. Other than this, the other functionalities are all the same as v0.0.7.
0.0.6
lax.select` generates a nan in the gradient from the inactive branch, while `lax.cond` doesn't