Interesting approach! Though, I am not quite sure why I need a middle-man library just to access the PyTorch library instead of performing the whole prediction in PyTorch?
Though it seems like you haven’t addressed other people’s complaints about the code readability. Therefore, I have created DataPredict Axon for easier debugging! I am already done with the most of the stuff already and am debating whether or not to add more features or focus on my master research thesis using the DataPredict library plus for consulting.