PYTORCH_TEST_WITH_ROCM=1 python test/nn/test_convolution.py --verbose
 ----------------------------------------------------------------------
Ran 1164 tests in 375.355s

OK (skipped=422, expected failures=25)

PYTORCH_TEST_WITH_ROCM=1 python test/test_nn.py --verbose
----------------------------------------------------------------------
FAIL: test_rnn_check_device (__main__.TestNN.test_rnn_check_device)
----------------------------------------------------------------------
RuntimeError: Expected all tensors to be on the same device, but got weight is on cpu, different from other tensors on cuda:0 (when checking argument in method wrapper_CUDA__miopen_rnn)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/var/lib/jenkins/pytorch/torch/testing/_internal/common_utils.py", line 3241, in wrapper
    method(*args, **kwargs)
  File "/var/lib/jenkins/pytorch/test/test_nn.py", line 3958, in test_rnn_check_device
    with self.assertRaisesRegex(RuntimeError,
AssertionError: "Input and parameter tensors are not at the same device" does not match "Expected all tensors to be on the same device, but got weight is on cpu, different from other tensors on cuda:0 (when checking argument in method wrapper_CUDA__miopen_rnn)"

To execute this test, run the following from the base repo dir:
    PYTORCH_TEST_WITH_ROCM=1 python test/test_nn.py TestNN.test_rnn_check_device

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0
----------------------------------------------------------------------
Ran 3634 tests in 476.463s

FAILED (failures=1, skipped=984, expected failures=3)