BUG: pd.api.types.is_scalar reports True for Troch tensors

[shwina]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd
import torch
pd.api.types.is_scalar(torch.tensor([0, 1, 2]))  # returns True

Issue Description

is_scalar reports True for torch tensors (of any dimensionality as far as I can tell).

Expected Behavior

I would expect this function to return False for torch tensors.

Installed Versions

In [4]: pd.version
Out[4]: '2.0.0'

In [5]: pd.show_versions()
/nvme/0/pgali/envs/cudfdev/lib/python3.10/site-packages/_distutils_hack/init.py:33: UserWarning: Setuptools is replacing distutils.
warnings.warn("Setuptools is replacing distutils.")

INSTALLED VERSIONS

commit : 478d340
python : 3.10.10.final.0
python-bits : 64
OS : Linux
OS-release : 4.15.0-76-generic
Version : #86-Ubuntu SMP Fri Jan 17 17:24:28 UTC 2020
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.0.0
numpy : 1.23.5
pytz : 2023.3
dateutil : 2.8.2
setuptools : 67.6.1
pip : 23.1
Cython : 0.29.34
pytest : 7.3.1
hypothesis : 6.72.0
sphinx : 5.3.0
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.12.0
pandas_datareader: None
bs4 : 4.12.2
bottleneck : None
brotli :
fastparquet : None
fsspec : 2023.4.0
gcsfs : None
matplotlib : None
numba : 0.56.4
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 11.0.0
pyreadstat : None
pyxlsb : None
s3fs : 2023.4.0
scipy : 1.10.1
snappy :
sqlalchemy : 1.4.46
tables : None
tabulate : 0.9.0
xarray : None
xlrd : None
zstandard : None
tzdata : 2023.3
qtpy : None
pyqt5 : None

[mroeschke]

Here are the "scalar" checks

pandas/pandas/_libs/lib.pyx

Lines 221 to 243 in 433f34b

	if (cnp.PyArray_IsAnyScalar(val)
	# PyArray_IsAnyScalar is always False for bytearrays on Py3
	or PyDate_Check(val)
	or PyDelta_Check(val)
	or PyTime_Check(val)
	# We differ from numpy, which claims that None is not scalar;
	# see np.isscalar
	or val is C_NA
	or val is None):
	return True
	# Next use C-optimized checks to exclude common non-scalars before falling
	# back to non-optimized checks.
	if PySequence_Check(val):
	# e.g. list, tuple
	# includes np.ndarray, Series which PyNumber_Check can return True for
	return False
	# Note: PyNumber_Check check includes Decimal, Fraction, numbers.Number
	return (PyNumber_Check(val)
	or is_period_object(val)
	or is_interval(val)
	or is_offset_object(val))

Debugging locally, it's interesting/odd that PySequence_Check returns False and PyNumber_Check returns True for a torch.Tensor. I would expect this to be the opposite for both checks

[mroeschke]

Opened up pytorch/pytorch#99646

[akashdasp]

is this is the issue with the pandas library it has the issue with the torch?

[mroeschke]

So it seems that from pytorch/pytorch#99646 that this is intended/won't fix behavior, and I don't think pandas can special case torch objects in this function as torch isn't a related dependency in pandas.

At best, maybe pandas could document that this is function is only reliable for python stdlib, numpy and pandas objects only?