Group G,Group D

[harishsdev]

test_five_crop
test_ten_crop
test_max_value
test_linear_transformation
test_autoaugment
test_center_crop
test_center_crop_2
test_color_jitter

[harishsdev]

PR for :#3945

[NicolasHug]

Thanks @harishsdev , there are failures, some of theme being undefined names: https://app.circleci.com/pipelines/github/pytorch/vision/8695/workflows/230c01ae-5eaa-4a88-9d4e-e480d9be8dfc/jobs/631032

Did you manage to run the tests locally?

[harishsdev]

@NicolasHug Thanks for your support,I'm able run tests in my local PC,I found no error there,

Please help to resolve below error

assertEqual(output_pil.size()[1:3], crop_size,
"image_size: {} crop_size: {}".format(input_image_size, crop_size))
E NameError: name 'assertEqual' is not defined

test/test_transforms.py:1948: NameError

[harishsdev]

My doubt is how to write assertEqual for formating expression!

[NicolasHug]

I'm able run tests in my local PC,I found no error there,

There's probably something wrong then because otherwise you would have gotten the same issue as above. did you run pytest test/test_transforms.py ?

My doubt is how to write assertEqual for formating expression!

you can juste write assert a == b, without an error message.

[harishsdev]

one function is failing as per my observation

pytest test/test_transforms.py -vvv -k test_five_crop-passed
pytest test/test_transforms.py -vvv -k test_ten_crop-passed
pytest test/test_transforms.py -vvv -k test_linear_transformation-failed
pytest test/test_transforms.py -vvv -k test_autoaugment -passed
pytest test/test_transforms.py -vvv -k test_center_crop -passed
pytest test/test_transforms.py -vvv -k test_center_crop_2 -passed
pytest test/test_transforms.py -vvv -k test_color_jitter-passed

[harishsdev]

Group G,Group D

PR for #3945

[harishsdev]

@NicolasHug please give me inputs if any further modifications required

[NicolasHug]

Thanks @harishsdev , we can parametrize a bit further but this looks quite good already

[NicolasHug]

we could parametrize over dtype here

[NicolasHug]

we can parametrize over single_dim

[NicolasHug]

let's parametrize over these 3 variables as well.

[NicolasHug]

to parametrize over input_images_sizes and preserve the current logic, we can do something like:

@pytest.mark.parametrize('odd_size', (True, False))
def test_center_crop_2(odd_size, ...):
    input_image_size = (random.randint(10, 32) * 2, random.randint(10, 32) * 2)
    if odd_size:
        input_image_size = (input_image_size[0] + 1, input_image_size[1] + 1)

[harishsdev]

(myenv) PS D:\contrib\19283975\exp\vision> pytest test/test_transforms.py -vvv -k test_five_crop
==================================================== test session starts ====================================================
platform win32 -- Python 3.8.8, pytest-6.2.3, py-1.10.0, pluggy-0.13.1 -- G:\installconda\python.exe
cachedir: .pytest_cache
rootdir: D:\contrib\19283975\exp\vision
plugins: anyio-2.2.0
collected 0 items / 1 error

========================================================== ERRORS ===========================================================
_________________________________________ ERROR collecting test/test_transforms.py __________________________________________
test\test_transforms.py:25: in
from common_utils import cycle_over, int_dtypes, float_dtypes
test\common_utils.py:20: in
from _assert_utils import assert_equal
test_assert_utils.py:11: in
assert_equal = functools.partial(torch.testing.assert_close, rtol=0, atol=0)
E AttributeError: module 'torch.testing' has no attribute 'assert_close'
===================================================== warnings summary ======================================================
G:\installconda\lib\site-packages\pyreadline\py3k_compat.py:8
G:\installconda\lib\site-packages\pyreadline\py3k_compat.py:8: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3, and in 3.9 it will stop working
return isinstance(x, collections.Callable)

-- Docs: https://docs.pytest.org/en/stable/warnings.html
================================================== short test summary info ==================================================
ERROR test/test_transforms.py - AttributeError: module 'torch.testing' has no attribute 'assert_close'
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Interrupted: 1 error during collection !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
================================================ 1 warning, 1 error in 1.59s ================================================

[NicolasHug]

I already replied to you about this issue here: #3975 (comment)

[harishsdev]

@NicolasHug Looks like all the tests passed finally,Can you please help me to merge this change

[NicolasHug]

Thanks @harishsdev , I took care of the prameterization of test_center_crop_2. I'll merge when green