benchmark

195efd50 - update template file; add optimize_for_inference (#378)

update template file; add optimize_for_inference (#378) * update template file; add optimize_for_inference * address Xu's feedback * try freezing * switch back to inference * print torch version * remove print version