You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
There are many complaints about passing scalar arguments to MXNet's ndarray calls (dmlc/minpy#139, dmlc/minpy#134). What I am thinking about is to have a wrapper for every registered NDArray function in MXNet such that if the given argument is scalar but supposed to be array, do an automatic conversion. My question is should we do it on python side? or on C++ side?
If done on python side, each ndarray call will have an extra wrapper (another function redirect), which may introduce overhead.
I don't know whether this is supported on C++ side, since I guess C++ cannot have arguments of either NDArray or scalar type?
There are many complaints about passing scalar arguments to MXNet's ndarray calls (dmlc/minpy#139, dmlc/minpy#134). What I am thinking about is to have a wrapper for every registered NDArray function in MXNet such that if the given argument is scalar but supposed to be array, do an automatic conversion. My question is should we do it on python side? or on C++ side?
@piiswrong @hotpxl @lryta
The text was updated successfully, but these errors were encountered: