I would like to deploy my PyTorch model (written in usual “research friendly eager mode” with Python API) within a signal processing framework (written in C code) and do inference.
What are the steps I should take to make it happen?
Basically what I want is:
Signal Processing Block (C code) --> Pytorch Model (takes in signal and outputs signal) --> Signal Processing Block (C code).
Is C a strong requirement or is cpp ok?
If you can do cpp, then you can use libtorch and the cpp api to run a jit model without python. This doc should be a good introduction about this.