Nightly test fail with "test_large_vector.test_nn" on v1.7.x branch

[ciyongch]

Description

Nightly test fail with "test_large_vector.test_nn" on v1.7.x branch, please check log here

Error Message

05:29:54  + nosetests-3.4 tests/nightly/test_large_vector.py:test_nn
05:44:46  E
05:44:46  ======================================================================
05:44:46  ERROR: test_large_vector.test_nn
05:44:46  ----------------------------------------------------------------------
05:44:46  Traceback (most recent call last):
05:44:46    File "/usr/local/lib/python3.5/dist-packages/nose/case.py", line 198, in runTest
05:44:46      self.test(*self.arg)
05:44:46    File "/work/mxnet/tests/nightly/test_large_vector.py", line 159, in test_nn
05:44:46      check_layer_norm()
05:44:46    File "/work/mxnet/tests/nightly/test_large_vector.py", line 89, in check_layer_norm
05:44:46      mx_out = nd.LayerNorm(data, gamma, beta, axis, eps)
05:44:46    File "<string>", line 81, in LayerNorm
05:44:46    File "/work/mxnet/python/mxnet/_ctypes/ndarray.py", line 91, in _imperative_invoke
05:44:46      ctypes.byref(out_stypes)))
05:44:46    File "/work/mxnet/python/mxnet/base.py", line 246, in check_call
05:44:46      raise get_last_ffi_error()
05:44:46  mxnet.base.MXNetError: MXNetError: Shape inconsistent, Provided = [4300000000], inferred shape=[5032704]
05:44:46  
05:44:46  ----------------------------------------------------------------------
05:44:46  Ran 1 test in 896.356s
05:44:46  
05:44:46  FAILED (errors=1)

@ChaiBapchya @szha @TaoLv

[ChaiBapchya]

layer norm and batchnorm have been fixed in master by #17805

But they aren't cherry-picked in 1.7.x as of today.
Cherry-picked them now and tested locally
Verified the large tensor tests pass

[ciyongch]

Hi @ChaiBapchya , thanks for your kindly help to identify and point out the fix:)

[ChaiBapchya]

http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/NightlyTestsForBinaries/detail/v1.7.x/4/pipeline
Nightly now passes. Fixed by #18261

[ciyongch]

Close this issue as it's fixed now.