GitLab

* Add attention mask and pad token warning to many of the models

* Remove changes under examples/research_projects

These files are not maintained by HG.

* Skip the warning check during torch.fx or JIT tracing

* Switch ordering for the warning and input shape assignment

This ordering is a little cleaner for some of the cases.

* Add missing line break in one of the files