test_loss.test_squared_hinge_loss has fixed seed that can mask flakiness

[szha]

The unit test in title have been using fixed seed to mask flakiness. Suggested action:

1. Evaluate whether the test is flaky without fixed seed. If not, remove seed. Else move to 2
2. If test is flaky, determine whether it's an actual uncaught edge case. If so, fix the operator. Else move to 3
3. If numerical instability is inevitable, adjust tolerance level appropriately.

[apeforest]

Thanks for filing this issue. We will investigate this Flaky test

[zachgk]

Ran tests/python/unittest/test_loss.py:test_squared_hinge_loss and tests/python/gpu/test_operator_gpu.py:test_squared_hinge_loss 10k times each and could not reproduce flakiness
Raised a PR #12017 to remove the fixed seed

CPU Run:

ubuntu@ip-172-31-27-232:~/incubator-mxnet$ MXNET_TEST_COUNT=10000 nosetests --logging-level=DEBUG --verbose -s tests/python/unittest/test_loss.py:test_squared_hinge_loss

[DEBUG] 9999 of 10000: Setting test np/mx/python random seeds, use MXNET_TEST_SEED=711236678 to reproduce.
[DEBUG] 10000 of 10000: Setting test np/mx/python random seeds, use MXNET_TEST_SEED=1402836224 to reproduce.
ok

----------------------------------------------------------------------
Ran 1 test in 7828.158s

OK

GPU Run:

ubuntu@ip-172-31-27-232:~/incubator-mxnet$ MXNET_TEST_COUNT=10000 nosetests --logging-level=DEBUG --verbose -s tests/python/gpu/test_operator_gpu.py:test_squared_hinge_loss

[DEBUG] 9999 of 10000: Setting test np/mx/python random seeds, use MXNET_TEST_SEED=330147794 to reproduce.
[DEBUG] 10000 of 10000: Setting test np/mx/python random seeds, use MXNET_TEST_SEED=608869560 to reproduce.
ok

----------------------------------------------------------------------
Ran 1 test in 8022.626s

OK

[haojin2]

@sandeep-krishnamurthy Related PR is merged, should be good to close.