Force the dtype of data to be float32 for test_correlation

[anirudh2290]

Description

Force the dtype of img1 and img2 to be float32. This should fix the intermittent breaking of test: test_correlation.

Please see: #9553

@eric-haibin-lin @marcoabreu

Checklist

Essentials

• Passed code style checking (make lint)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage:
• Unit tests are added for small changes to verify correctness (e.g. adding a new operator)
• Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore)
• Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL)
• Code is well-documented:
• For user-facing API changes, API doc string has been updated.
• For new C++ functions in header files, their functionalities and arguments are documented.
• For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
• To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change

[szha]

How is the fix verified? Do you have an idea why it started failing?

[eric-haibin-lin]

@szha Yes, we found a np seed/state that consistently fails the test. The random np input is in float64 dtype while the MXNet ndarray is using float32. Hence the difference when comparing the result.

[marcoabreu]

So am I right when I assume that this problem could also impact other tests because of a dtype mismatch? Maybe we should need some type of general validation? Am 26.01.2018 9:19 nachm. schrieb "Haibin Lin" <notifications@github.com>:

…

Merged #9581 <#9581>. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#9581 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ARxB6-zic3WAs6iH3YypgkdvQJNcORQ7ks5tOrHxgaJpZM4RvBFi> .

[eric-haibin-lin]

@marcoabreu yes this could be a common problem to all other unit test using np random inputs