Under this repo we have a cli folder and a server folder; within server there is a Go server and under /python a folder called torchserver which holds code for a server that works as a pythorch wrapper.
The cli folder holds the command line client which can be used to create a definition file and deploy code -- it serializes the code and sends it to the server which in turn would deserialize it and pass it to the torchserver to expose models that our Go server receives and loads. This step is lacking as there are bugs to be ironed out and I had to put this project to a halt due to time restrictions.
current implementation just serves as a toy to communicate cli and server.