[Python] Cupy throwing a unhashable type: 'ndarray' when appending nested tuples to a set

For a project we are trying to convert numpy code to cupy code for the speedup but we are getting this error:

TypeError: unhashable type: 'ndarray'


When we run

import cupy as cp
newMat = cp.array([[ 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.], [ 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., -1.]])
newMatSet = set()
newMatSet.add(tuple(tuple(row) for row in sortBis(newMat)))


With sortBis simply reordering the matrix:

def sortBis(mat: np.ndarray):
colInds = cp.lexsort(mat[:, 1:])
mat[:, 1:] = mat[:, 1:][:, colInds]
return mat


It is my understanding that a nested tuple should be hashable, and this code works perfectly find with numpy. Does anyone know of a fix?

Continue reading...