💊 CI failures summary and remediations

As of commit 0fed3ed (more details on the Dr. CI page):

• 1/2 failures possibly* introduced in this PR
  • 1/1 non-CircleCI failure(s)
• 1/2 tentatively recognized as flaky ❄️
  • Click here to rerun these jobs

❄️ 1 failure tentatively classified as flaky

but reruns have not yet been triggered to confirm:

pytorch_linux_xenial_cuda10_2_cudnn7_py3_gcc7_test2 (1/1)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun) ❄️

Jan 14 05:43:21                        ~~~~ <--- HERE
Jan 14 05:43:21 RuntimeError: CUDA error: an illegal memory access was encountered
Jan 14 05:43:21 
Jan 14 05:43:21 
Jan 14 05:43:21 ======================================================================
Jan 14 05:43:21 ERROR [0.203s]: test_where_and_typing (__main__.TestTEFuser)
Jan 14 05:43:21 ----------------------------------------------------------------------
Jan 14 05:43:21 Traceback (most recent call last):
Jan 14 05:43:21   File "test_jit_fuser_te.py", line 1142, in test_where_and_typing
Jan 14 05:43:21     x = torch.randn(4, 4, dtype=torch.double, device=device)
Jan 14 05:43:21 RuntimeError: CUDA error: an illegal memory access was encountered
Jan 14 05:43:21 
Jan 14 05:43:21 ======================================================================
Jan 14 05:43:21 ERROR [0.176s]: test_zero_element_tensors_cuda (__main__.TestTEFuser)
Jan 14 05:43:21 ----------------------------------------------------------------------
Jan 14 05:43:21 Traceback (most recent call last):
Jan 14 05:43:21   File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_utils.py", line 888, in wrapper
Jan 14 05:43:21     method(*args, **kwargs)
Jan 14 05:43:21   File "test_jit_fuser_te.py", line 178, in test_zero_element_tensors_cuda
Jan 14 05:43:21     self._test_zero_element_tensors(device="cuda")
Jan 14 05:43:21   File "test_jit_fuser_te.py", line 174, in _test_zero_element_tensors

ci.pytorch.org: 1 failed

• Failed: pr/pytorch-linux-bionic-rocm3.10-py3.6

I suppose it's worth asking, should we also rename is_deterministic?

Also, this PR only changes the Python torch.set_deterministic function, it does not change the C++ interface. Should I change the C++ as well?

Hey @mruberry, the name we chose, torch.is_deterministic_algorithms_enabled(), has been bothering me a little bit because it's not grammatically correct :P What do you think about torch.are_deterministic_algorithms_enabled()? I get that "is" is a much more common term for boolean state checking functions, but in this case, I think "are" feels more natural and could be easier for users to remember. What do you think?

Hey @mruberry, the name we chose, torch.is_deterministic_algorithms_enabled(), has been bothering me a little bit because it's not grammatically correct :P What do you think about torch.are_deterministic_algorithms_enabled()? I get that "is" is a much more common term for boolean state checking functions, but in this case, I think "are" feels more natural and could be easier for users to remember. What do you think?

Sounds great.

@mruberry , this PR is ready for another review. The ROCm failure doesn't seem to be related

@mruberry merged this pull request in 8ab1a14.