Add dtype and manipulation functions by npolina4 · Pull Request #1566 · IntelPython/dpnp

npolina4 · 2023-09-21T16:18:02Z

RP will separate:
#1582 Leverage dpctl.tensor.iinfo() and dpctl.tensor.finfo() implementation.
#1583 Clean up old implementation of logic functions
#1585 Updated atleast_2d and atleast_3d functions
#1586 Implemented dpnp.tile function.

Add dtype and manipulation functions and tests.
Manipulation functions: dpnp.broadcast_arrays, dpnp.repeat, dpnp.vstack, dpnp.ravel
Data Type functions: dpnp.can_cast

Have you provided a meaningful PR description?
Have you added a test, reproducer or referred to 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?

github-actions · 2023-09-22T19:38:41Z

View rendered docs @ https://intelpython.github.io/dpnp/pull/1566/index.html

vtavana · 2023-09-25T16:38:49Z

It seems we do not to skip the fallowing tests anymore.

tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_0_{shapes=[(), ()]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_1_{shapes=[(0,), (0,)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_2_{shapes=[(1,), (1,)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_3_{shapes=[(2,), (2,)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_4_{shapes=[(0,), (1,)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_5_{shapes=[(2, 3), (1, 3)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_6_{shapes=[(2, 1, 3, 4), (3, 1, 4)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_7_{shapes=[(4, 3, 2, 3), (2, 3)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_8_{shapes=[(2, 0, 1, 1, 3), (2, 1, 0, 0, 3)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_9_{shapes=[(0, 1, 1, 3), (2, 1, 0, 0, 3)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestBroadcast_param_10_{shapes=[(0, 1, 1, 0, 3), (5, 2, 0, 1, 0, 0, 3), (2, 1, 0, 0, 0, 3)]}::test_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestInvalidBroadcast_param_0_{shapes=[(3,), (2,)]}::test_invalid_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestInvalidBroadcast_param_1_{shapes=[(3, 2), (2, 3)]}::test_invalid_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestInvalidBroadcast_param_2_{shapes=[(3, 2), (3, 4)]}::test_invalid_broadcast_arrays
tests/third_party/cupy/manipulation_tests/test_dims.py::TestInvalidBroadcast_param_3_{shapes=[(0,),(2,)]}::test_invalid_broadcast_arrays

ndgrigorian · 2023-10-05T18:44:04Z

@npolina4
With IntelPython/dpctl#1427 you will no longer need to flatten the repeated array for the axis=None case. Array API spec seems to be supporting this behavior.

antonwolfy · 2023-10-09T13:02:11Z

The below example works differently with numpy and description:

a=numpy.ones((2, 3), dtype='i4')
b=a.astype('i4', copy=False)
bisa# Out: Truea=dpnp.ones((2, 3), dtype='i4')
b=a.astype('i4', copy=False)
bisa# Out: False

antonwolfy · 2023-10-09T12:38:23Z

+ Parameters
+ ----------
+ dtype : dtype
+ Target data type.


What is about None value? It would be good to state clearly, since dpctl and numpy supports that but might behave differently.

antonwolfy · 2023-10-09T12:44:55Z

- return dpt.astype(x1, dtype, order=order, casting=casting, copy=copy)
+ Parameters
+ ----------
+ dtype : dtype


Need to describe x1 and casting

antonwolfy · 2023-10-09T12:46:37Z

- Otherwise, a copy is returned.
+ Parameters
+ ----------
+ dtype : dtype


Need to describe casting

antonwolfy · 2023-10-09T12:48:24Z

+ Limitations
+ -----------
+ Parameter `subok` is supported with default value.
+ Otherwise the function will be executed sequentially on CPU.


Probably it would be better to get rid of the falling back on CPU and to raise an exception explicitly, like cupy does?

antonwolfy · 2023-10-09T13:00:21Z

- new_array = self.__new__(dpnp_array)
- new_array._array_obj = dpt.astype(
- self._array_obj, dtype, order=order, casting=casting, copy=copy
+ return dpnp.astype(


It would be better to have main implementation here and then dpnp.astype will be equal to
if dpnp.is_supported_array_type(x1): return x1.astype(...) else: raise TypeError(...)
Also it will allow us to avoid invoking dpnp.get_usm_ndarray(...) and dpnp_array._create_from_usm_ndarray(...).

antonwolfy · 2023-10-09T13:31:25Z

@@ -150,30 +151,41 @@ def asnumpy(input, order="C"):


 def astype(x1, dtype, order="K", casting="unsafe", subok=True, copy=True):


Neither numpy or cupy has astype method (they have it only in the context of ndarray), so I would propose to remove unsupported subok parameter here.

antonwolfy · 2023-10-09T13:36:22Z

+ -------
+ out : dpnp.array
+ If ``copy`` is False and no cast is required, then the array itself is returned.
+ Otherwise, it returns a (possibly casted) copy of the array.


It sounds a bit unclear what "possibly casted" means. Could you please clarify that in the description or rephrase it.
In general, numpy and dpctl returns a view where possible.

antonwolfy · 2023-10-09T13:45:11Z


 def astype(self, dtype, order="K", casting="unsafe", subok=True, copy=True):
- """Copy the array with data type casting.
+ """


Could you please check tests/third_party/cupy/core_tests/test_ndarray_copy_and_view.py::TestArrayCopyAndView and unmute where possible?
Also it would be great refresh test_ndarray_copy_and_view.py file.

npolina4 added 2 commits September 19, 2023 13:28

Add dtype functions
fabf886

Add atleast 2d and 3d, vstack, repeat, and broadcast_arrays functions
c02823f

npolina4 requested a review from vlad-perevezentsev September 21, 2023 16:18

vtavana reviewed Sep 21, 2023
View reviewed changes

npolina4 added 2 commits September 22, 2023 13:54

address comments
f6738ff

Merge branch 'master' into dtype_functions
1faf3cc

npolina4 added 2 commits September 22, 2023 17:02

Add more tests
86e052c

Skip test with random on Iris Xe
c442a9e

npolina4 added 2 commits September 25, 2023 12:56

Removed broadcast_arrays tests from skip list
fa06664

Merge branch 'master' into dtype_functions
de1788b

vlad-perevezentsev reviewed Sep 25, 2023
View reviewed changes

Comment threaddpnp/dpnp_iface_types.py

Clean up dpnp_astype and dpnp_flatten functions
485fb60

vtavana reviewed Sep 29, 2023
View reviewed changes

Comment threaddpnp/dpnp_algo/dpnp_algo.pxd

vtavana reviewed Sep 29, 2023
View reviewed changes

Comment threaddpnp/dpnp_array.py

Old implementation removed
201a8a3

vtavana reviewed Oct 3, 2023
View reviewed changes

Comment threaddpnp/dpnp_iface_manipulation.py Outdated
Comment threaddpnp/dpnp_iface_manipulation.py Outdated
Comment threaddpnp/dpnp_iface_manipulation.py Outdated
Comment threaddpnp/dpnp_algo/dpnp_algo.pxd

npolina4 added 6 commits October 5, 2023 11:31

address comments
589f947

Merge branch 'master' into dtype_functions
cb1b909

fix pre-commit
5164a7b

fix pre-commit
ab8e482

fix pre-commit
9182767

fix pre-commit
30a3434

vtavana approved these changes Oct 6, 2023
View reviewed changes

Merge branch 'master' into dtype_functions
df4bbfa

antonwolfy requested changes Oct 9, 2023
View reviewed changes

npolina4 closed this Nov 28, 2023

npolina4 deleted the dtype_functions branch November 28, 2023 15:55

		@@ -150,30 +151,41 @@ def asnumpy(input, order="C"):


		def astype(x1, dtype, order="K", casting="unsafe", subok=True, copy=True):

Uh oh!

Conversation

npolina4 commented Sep 21, 2023• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actionsBot commented Sep 22, 2023

Uh oh!

vtavana commented Sep 25, 2023

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ndgrigorian commented Oct 5, 2023

Uh oh!

antonwolfy commented Oct 9, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

npolina4 commented Sep 21, 2023•
edited
Loading