Hello everybody,
As contributor to Apache MXNet project I would like to ask community a
couple of questions in regards to C Predict and C++ APIs that MXNet
provides to its users. My main goal is to better understand the pain points
community members currently see/have with those APIs as well as to what
contributions to C++ and C Predict APIs would be most beneficial to users
who are using/tried to use these APIs of Apache MXNet.
1. What are the typical uses cases in which C++ (cpp-package) or C (C
Predict) APIs are used? For example: inference, training or both.
2. Which set of APIs out of C++ and C do users prefer? Preferably
with reasons why.
3. What are the frequently used platforms (Linux, Mac, Windows, etc)
and configurations (such as CPU, GPU, etc) on which these APIs are used?
4. Currently, users are required to build these APIs from source.
Would it be helpful if these APIs are available as standalone packages
distributed via package managers (example: apt-get)?
I would highly appreciate your replies to any or all of the above questions.
Thanks,
-Amol