Fix validation transforms in ResNet example

[Yancey1989]

The transforms on the validation dataset are different from training data. c.f. https://github.com/pytorch/examples/blob/master/imagenet/main.py#L225

[codecov[bot]]

Codecov Report

Merging #365 into develop will not change coverage. The diff coverage is n/a.

@@           Coverage Diff            @@
##           develop     #365   +/-   ##
========================================
  Coverage    87.62%   87.62%           
========================================
  Files           33       33           
  Lines         1487     1487           
========================================
  Hits          1303     1303           
  Misses         121      121           
  Partials        63       63           

[shendiaomo]

The transforms on the validation dataset are different from training data. c.f. https://github.com/pytorch/examples/blob/master/imagenet/main.py#L225

Great. This may lead to unstable accuracy on the validation dataset.