Fixing AttributeError in slice_indicator initialization

[rafe-sh]

The issue was related to an AttributeError raised in the code when initializing the slice_indicator variable. The error occurred due to the usage of np.int as the dtype argument in the np.ones() function. In earlier versions of NumPy, np.int was a deprecated alias for the built-in int type. However, starting from NumPy 1.20, these aliases were removed.

To resolve the issue, the code was modified by changing the dtype argument from np.int to np.int64 to specify the desired precision explicitly. The modified code snippet looks as follows:

# turn partitions into an indicator
slice_indicator = np.ones(y.shape[0], dtype=np.int64)

This change ensures compatibility with the latest versions of NumPy and resolves the AttributeError related to the use of np.int in the code. The modified code should now execute without any errors.

By making this fix, the issue has been resolved, and the code functions as intended, initializing the slice_indicator variable with the correct dtype.

[gunsodo]

@rafe-sh I noticed there is another np.int in sliced.datasets as well. https://github.com/joshloyal/sliced/blob/243bde236d8c615f9563279a0a2095e2fa2f4650/sliced/datasets/base.py#L17

@joshloyal May I kindly ask if there is any plan for releasing a new version including this fix that @rafe-sh has made? Anyway, I sincerely thank you for making this repository!

[statmlben]

A quick fix for this error is to add one line before using sliced:

import numpy as np
np.int = np.int64

Of course, this is an ad-hoc solution. Hope the authors can merge this commit soon.