The float version of calc_digamma should return float type. (#25488)
Summary:
Besides common understanding, the only occurrence of calc_digamma is in UnaryOpsKernel.cpp, which clearly sees the float version of calc_digamma as returning float type (and the double version of calc_digamma as returning double type).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25488
Reviewed By: ezyang
Differential Revision: D17172379
Pulled By: VitalyFedyunin
fbshipit-source-id: 56facd45564cff019d572138c0d541a0bdded5c8