In GPJax/gpjax/kernels.base.py, the return type of the diagonal method in AbstractKernel should change from
def diagonal(self, x: Num[Array, "N D"]) -> Float[Array, " N"]:
to
def diagonal(self, x: Num[Array, "N D"]) -> LinearOperator:
to both match the return type of gram() and the expected return type in AbstractKernelComputation.