It seems that to make full use of TensorFlow and its associated libraries, I need to access it with Python. Using TensorFlow requires creating fairly complex graphs that automate complex Python helper libraries (such as Keras) in complex ways. I see no way to do this directly from .NET. See, for example, the relevant GitHub discussion on C # support .
My idea is this: the main C # application is called in a Python application to access TensorFlow. That way, most of the code will remain in C # land. I can use CPython, which fully supports everything (IronPython does not).
How can I "invoke a Python application"? I assume that I will need to run python.exe in a child process and make it execute commands. The child process must remain between the teams so that I can store / exchange objects and data.
How am I best to do this? I don’t quite know where to start.
source share