is there a way to seperate initialization of wavenet and run?

[isaacleeai]

When using wavenet through nv_wavenet.py, everytime one wants to run an inference, one has to initialize the gpu-side wavenet and run the inference. Is there a way to initialize the structure only once and run inference multiple times, while still using python to initiate inference?