When deploy PyMC3 model, must we deploy g++ to get less running time?


#1

Questions:
In my control device, I can’t deploy g++ because of some reasons. But if I can’t deploy g++, there is a warning " WARNING (theano.configdefaults): g++ not detected! Theano will be unable to execute optimized C-implementations (for both CPU and GPU) and will default to Python implementations. Performance will be severely degraded." .
Must I deploy g++ to get less running time? Why need pymc3 a compiler? Thank you.


#2

because theano use g++ to compile the model in c code so it runs a bit faster. However, if you are deploying the model, maybe there is a way run the model once where g++ is available, and then copy the generated c code somewhere.
I have never actually done it, so I am not sure if it works or not.


#3

Thank you. but if for different case, it needs to be instantiated. So it may not be same c code.