szha commented on a change in pull request #8377: re-enable tests URL: https://github.com/apache/incubator-mxnet/pull/8377#discussion_r147319165
########## File path: tests/python/gpu/test_operator_gpu.py ########## @@ -1359,17 +1359,37 @@ def test_rnn_layer(): def test_sequence_reverse(): check_sequence_reverse(mx.gpu(0)) -@unittest.skip("Test fails intermittently. Temporarily disabled until fixed. Tracked at https://github.com/apache/incubator-mxnet/issues/8211") def test_autograd_save_memory(): - x = mx.nd.zeros((128, 512, 512), ctx=mx.gpu(0)) + """test_autograd_save_memory. + Flaky, may cause OOM. Tracked at https://github.com/apache/incubator-mxnet/issues/8211""" + import platform + + def get_mem(): Review comment: This is to test whether autograd memory-save fix is working or not. Given that our CI slaves haven't been homogenous (of which the reason eludes me), the hard-coded number may result in a max memory usage above available mem. So after this change, instead of testing whether max number causes OOM, it tests whether the increment is within limit. I think this test being platform-dependent is OK since the path being tested is not necessarily platform dependent. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services