To avoid problems with metadata in dtypes when copying hdf5 arrays to zarr or npy file or in other cases: - https://github.com/zarr-developers/zarr-python/issues/1152 - https://github.com/numpy/numpy/issues/15488 - https://github.com/numpy/numpy/issues/23169 i add a numpy tool function that creates a dtype from an original one but without any metada inside. https://github.com/ninousf/numpy/tree/dtype-without-metadata what do you think of it ? thanks