Performing simple inference

[maxwellflitton]

Thanks for creating such a great library. I went through the math in the README and then created some simple data that would mimic the decision tree in the README and trained a basic DecisionTreeClassifier. After this, I got the state_dict of the model which in the debugger looked like the following:

Does this mean that if I have an input tensor, and I merely loop through the matrices performing matrix multiplications in order to get accurate outcomes? I'm asking this because I am looking into performing inference with as few dependencies as possible. Ideally, I'd like to build a compiled program in Rust that just loads those weights and performs the matrix multiplications to get predictions.