Add test to check if CuDNN is enabled #1715
Conversation
miguelgfierro
requested review from
gramhagen,
anargyri,
loomlike and
wutaomsft
as code owners
Codecov Report
@@ Coverage Diff @@
## staging #1715 +/- ##
============================================
+ Coverage 0.00% 23.33% +23.33%
============================================
Files 87 87
Lines 9132 9132
============================================
+ Hits 0 2131 +2131
- Misses 0 7001 +7001
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
| @@ -46,6 +46,11 @@ def test_get_cudnn_version(): | |||
| assert get_cudnn_version() > "7.0.0" | |||
|
|
|||
|
|
|||
| @pytest.mark.gpu | |||
| def test_cudnn_enabled(): | |||
| assert torch.backends.cudnn.enabled == True | |||
There was a problem hiding this comment.
assert torch.backends.cudnn.enabled would also work here.
There was a problem hiding this comment.
Thanks for reviewing Fatemeh
The problem with just using assert is that it also takes other values. For exampe if you have something like this:
def test_simple():
assert 1
assert True
assert "a"
assert False
you'll get 3 oks and one error.
There was a problem hiding this comment.
That is correct, in this scenario assert torch.backends.cudnn.enabled can only be True or False, so we still get the desired behavior in either cases. But I agree, your way is better for being more explicit and readable. Dropping ==True is only more concise.
Description
To make sure we have enabled in our test system cudnn.
This might help with the problem @pradnyeshjoshi is having
Related Issues
Checklist:
staging branchand not tomain branch.