Implements dpctl.tensor.clip

[ndgrigorian]

This pull request implements the function clip, for clipping elements of an array to a range.
Clip permits both Python scalar and dpctl.tensor.usm_ndarray arguments for its min and max arguments.

An example of the new functionality:

In [1]: import dpctl.tensor as dpt, numpy as np

In [2]: x = dpt.arange(-2, 8, dtype="i4")

In [3]: dpt.clip(x, 0, 5)
Out[3]: usm_ndarray([0, 0, 0, 1, 2, 3, 4, 5, 5, 5], dtype=int32)

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• If this PR is a work in progress, are you opening the PR as a draft?

[github-actions]

View rendered docs @ https://intelpython.github.io/dpctl/pulls/1444/index.html

[coveralls]

coverage: 85.728% (+0.01%) from 85.718% when pulling d839124 on implement-clip into 386bd8b on master.

[oleksandr-pavlyk]

Encountered TypeError:

In [1]: import dpctl, dpctl.tensor as dpt, dpctl.utils as du

In [2]: x = dpt.arange(10)

In [3]: dpt.clip(x, 0, 10)
Out[3]: usm_ndarray([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

In [4]: dpt.clip(x, 0, 5)
Out[4]: usm_ndarray([0, 1, 2, 3, 4, 5, 5, 5, 5, 5])

In [5]: dpt.clip(x, 0, dpt.inf)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_45 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[ndgrigorian]

Encountered TypeError:

In [1]: import dpctl, dpctl.tensor as dpt, dpctl.utils as du

In [2]: x = dpt.arange(10)

In [3]: dpt.clip(x, 0, 10)
Out[3]: usm_ndarray([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

In [4]: dpt.clip(x, 0, 5)
Out[4]: usm_ndarray([0, 1, 2, 3, 4, 5, 5, 5, 5, 5])

In [5]: dpt.clip(x, 0, dpt.inf)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)

This is currently the expected behavior, as the output is expected to be the same type as the input, and a float can't be cast to int.

[antonwolfy]

Checked with existing dpnp tests around clip and all of them are passed.

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_45 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_46 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_50 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_53 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[ndgrigorian]

@oleksandr-pavlyk
Do you see value in applying the change of TypeError to ValueError for, e.g., dpt.clip(dpt.arange(10, dtype="i8"), 0, dpt.inf) to the elementwise functions as well?

Then for instance dpt.bitwise_and(dpt.arange(10, dtype="i8"), dpt.inf) will also raise a ValueError instead of TypeError.

[oleksandr-pavlyk]

@ndgrigorian I think ValueError makes more sense in the example dpt.bitwise_and(dpt.arange(10, dtype="i8"), dpt.inf) as well. Thanks for bringing this up @ndgrigorian

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_70 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[ndgrigorian]

@ndgrigorian I think ValueError makes more sense in the example dpt.bitwise_and(dpt.arange(10, dtype="i8"), dpt.inf) as well. Thanks for bringing this up @ndgrigorian

Making this change was more breaking than expected for elementwise functions. It will belong in a separate PR, as it will have to touch a lot of tests and unary functions. There are also other cases where inappropriate dtype of an input array leads to a TypeError throughout the project. It may make sense to change those as well.

For now, that work is done on just clip. The PR is ready for review.

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_70 ran successfully.
Passed: 935
Failed: 65
Skipped: 119

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev0=py310ha25a700_70 ran successfully.
Passed: 934
Failed: 66
Skipped: 119