SamplingKind.block_sumsq

abstractmethod t3toolbox.backend.fitting.SamplingKind.block_sumsq(out, n_w)

def block_sumsq(self, out, n_w):

Per-(mode, order) |.|^2 -> a 2-D matrix; UNWEIGHTED (the honest diagnostic table).

Optional: only the Newton-CG relative-error display needs it. A kind that implements it must also set has_block_sumsq = True so the display knows (the two output-shape bases below do).