DistillingNN-SDT

I achieved 95.47% of accuracy on test set (MNIST dataset) on a 9-level tree after 23 epoches of training (without distilling) with hyperparameters reported on top of main.py (learned best weights in best_ckpt folder).

Details