Fixed * Issue in `_combine_mean_stds` when nums is less than 1 and bessel correction is on. * Added `__params__` to Categorical * Fixed case where `rng` passed to `Normal.random` was not respected. * Ensure `kwarray.group_indicies` has consistent output across numpy versions
0.6.13
Fixed * Bug in robust normalizers with quantile-clip extrema
0.6.12
Fixed: * Problems with last release
0.6.11
Changed * Avoiding global torch imports. This may slow down the ArrayAPI, and we should revisit a refactor after removing 3.6 support.
0.6.10
Added * Added `nan_policy` argument to `kwarray.Sticher` * Added `array_equal` to ArrayAPI
Changed * Allow `None` as `indices` in `kwarray.Sticher.add` * Change `nan_behavior` to `nan_policy` in `kwarray.RunningStats` * Updated `ub.repr2` to `ub.urepr`.
Fixed * `robust_normalize` now preserves masked arrays.