Function merge_padding_and_attention_mask does not return an output with the desired shape when both padding and attention masks are given

[abodinier]

Function merge_padding_and_attention_mask does not return an output with the desired shape when both padding and attention masks are given.

See the linked issue here: #783 (comment)

Currently, the TransformerEncoder layers breaks if we call it with both padding and attention masks.

Changes:

• Remove duplicated tf.newaxis
• Create a check function to control the shapes of the padding and attention masks
• Update unit tests to make sure that the error is raised if we give merge_padding_and_attention_mask masks with bad shapes.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[atharvapurdue]

Hi @abodinier , great job!
I think there is small change required. merged_mask variable is assigned but never used. The check should pass after that

[mattdangerw]

This looks great! Some minor comments, this also still has a few minor formatting errors (./shell/format.sh), and needs to be merged with master.

Thanks!