I had a brief look at your package, and my impression was that it's only changing model training. If this is correct then the format of the model.txt (calling `lgbm.save(model, "model.txt")`) is the same as regular lightgbm. This would mean you can use my library for inference.