Remove p-value checks in test_transforms.py #4756
Conversation
💊 CI failures summary and remediationsAs of commit efb94f8 (more details on the Dr. CI page):
1 failure not recognized by patterns:
1 job timed out:
This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
| @@ -1362,160 +1361,42 @@ def test_to_grayscale(): | |||
| trans4.__repr__() | |||
|
|
|||
|
|
|||
| @pytest.mark.skipif(stats is None, reason="scipy.stats not available") | |||
| def test_random_grayscale(): | |||
There was a problem hiding this comment.
I moved this one into test_randomness above. It will cover all the randomness-related checks in this test. For the rest (expected values), everything is already covered in test_to_grayscale() just above.
| @@ -2011,72 +1893,6 @@ def test_randomperspective_fill(mode): | |||
| F.perspective(img_conv, startpoints, endpoints, fill=tuple([fill] * wrong_num_bands)) | |||
|
|
|||
|
|
|||
| @pytest.mark.skipif(stats is None, reason="scipy.stats not available") | |||
| def test_random_vertical_flip(): | |||
There was a problem hiding this comment.
moved this one in test_randomness above, same for horizontal split.
| assert_equal(gray_np, gray_np_3) | ||
| @pytest.mark.parametrize("seed", range(10)) | ||
| @pytest.mark.parametrize("p", (0, 1)) | ||
| def test_random_apply(p, seed): |
There was a problem hiding this comment.
I could move this one up with test_randomness but this would require taking care of the case where fn is None -- this transform doesn't have a functional equivalent. Not worth it IMHO considering how simple the code is, but I don't mind
There was a problem hiding this comment.
Including this above will also have the negative side-effect of requiring you to modify your check condition for p=1 to out != img which IMO is less strong to what you have above. I agree keeping this as you have it.
NicolasHug
changed the title
| assert_equal(gray_np, gray_np_3) | ||
| @pytest.mark.parametrize("seed", range(10)) | ||
| @pytest.mark.parametrize("p", (0, 1)) | ||
| def test_random_apply(p, seed): |
There was a problem hiding this comment.
Including this above will also have the negative side-effect of requiring you to modify your check condition for p=1 to out != img which IMO is less strong to what you have above. I agree keeping this as you have it.
|
Hey @NicolasHug! You merged this PR, but no labels were added. The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py |
Summary: * Change test_random_apply * Change test_random_choice * Change test_randomness * took care of RandomVert/HorizFlip * take care of RandomGrayScale * minor cleanup * avoid 0 degree rotation just in case Reviewed By: datumbox Differential Revision: D32064690 fbshipit-source-id: 8a67a56fe1e959d144b2cdf49c76610e00933508
* Change test_random_apply * Change test_random_choice * Change test_randomness * took care of RandomVert/HorizFlip * take care of RandomGrayScale * minor cleanup * avoid 0 degree rotation just in case
Towards #4506 (comment)
This PR re-works most p_value tests in
test_transforms.py.There are still a few p_value tests in the file, but they're testing functionalities that aren't as simple as just doing
if torch.rand(1) < p: return orig_img else ..., so we can't simplify them in the same waycc @vfdev-5 @datumbox @pmeier