TypeError: bool() conversion fails on single-element arrays with ndim > 0

[abagusetty]

dpnp raises a TypeError when evaluating the truth value of a single-element array with shape (1, 1), while NumPy and CuPy both handle this correctly.

import dpnp

xy_norm = dpnp.ones((1, 1)) * 1.5

if xy_norm > 1e-14:  # TypeError raised here
    print("works")

Output:

  File "/home/abagusetty/gpu4pyscf-testing/gpu4pyscf/./test_dpnp_newissue_v1.py", line 6, in <module>
    if xy_norm > 1e-14:
       ^^^^^^^^^^^^^^^
  File "/home/abagusetty/gpu4pyscf-testing/dpnp/dpnp/dpnp_array.py", line 199, in __bool__
    return self._array_obj.__bool__()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "dpctl/tensor/_usmarray.pyx", line 1143, in dpctl.tensor._usmarray.usm_ndarray.__bool__
  File "dpctl/tensor/_usmarray.pyx", line 130, in dpctl.tensor._usmarray._check_0d_scalar_conversion
TypeError: only 0-dimensional arrays can be converted to Python scalars