[ROCm] Fix unit tests on CI

[iotamudelta]

Disables two of the unit tests in test_cuda that got introduced after test_cuda was enabled that fail on ROCm.

[facebook-github-bot]

ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[ssnl]

I'm really confused why these tests fail with

02:00:13 ======================================================================
02:00:13 ERROR: test_max_with_inf (__main__.TestCuda)
02:00:13 ----------------------------------------------------------------------
02:00:13 Traceback (most recent call last):
02:00:13   File "/var/lib/jenkins/workspace/test/common.py", line 251, in wrapper
02:00:13     method(*args, **kwargs)
02:00:13   File "test_cuda.py", line 1700, in test_max_with_inf
02:00:13     TestTorch._test_max_with_inf(self, (torch.half, torch.float, torch.double), 'cuda')
02:00:13   File "/var/lib/jenkins/workspace/test/test_torch.py", line 800, in _test_max_with_inf
02:00:13     self.assertTrue(torch.all(torch.max(a, dim=1)[0] == inf).item())
02:00:13 RuntimeError: Expected object of scalar type Half but got scalar type Long for argument #0 'max'

Any ideas?

[iotamudelta]

@ssnl unclear. We are working on unit test pass rate ATM. There are a few things that could interfere here: a) we are aware of a compiler bug (fixed in #11198 ) that can cause hangs and crashes, b) we are also aware of a few tests (min/max in particular) that succeed on our nodes but fail on the CI (we are looking into this).