Alvin Wan

@ccl-private Thanks for trying out the repo and posting here. Checking this out now, it looks like our scripts can't find Cadene's FC layers. I'll post here when I've got...

Hm are those the outputs after training the model twice? If so, double-check the script downloaded/loaded the pretrained WRN28x10 correctly. Or do you mean those are the outputs after training...

Thanks -- unfortunately, it does not support Tensorflow, but PRs are welcome!

@oak-tree Hope this helps: Model-related wrappers are in [nbdt/model.py](https://github.com/alvinwan/neural-backed-decision-trees/blob/master/nbdt/model.py). The data structure in [data/custom.py](https://github.com/alvinwan/neural-backed-decision-trees/blob/master/nbdt/data/custom.py) and utils are framework-agnostic though. Beyond that, unfortunately, I'm not familiar with Tensorflow so don't know...

Thanks for posting -- there isn't at the moment, unfortunately, but here's a reply that may be helpful: https://github.com/alvinwan/neural-backed-decision-trees/issues/13#issuecomment-765877801

Hey @MarkTensenSgt, Thanks for posting! Here's the structure for each `decision` in that output https://github.com/alvinwan/neural-backed-decision-trees/blob/fe22aefe99afce813d3afb4a5976946e33e844f0/nbdt/model.py#L175. There isn't any formal documentation unfortunately.

@eladmw Thanks for asking -- do you mean regression? If so, not directly. However, if each of the leaves has a dedicated (possibly-learned) continuous value like in CART, you could...

@XAVILLA Ah, good question! Do you have a regular pre-trained model for that dataset? The most stable way to train is to: 1. Train a regular model 2. Generate a...

Hey @wangdemo, It can indeed run on Windows, as there aren't system-dependent parts of the package and folks have previously done this https://github.com/alvinwan/neural-backed-decision-trees/pull/17. However, I don't own a Windows machine...

@Muzijiajian Hm, are you on PyTorch 1.4? https://github.com/alvinwan/neural-backed-decision-trees/blob/master/requirements.txt#L2. Your code looks right, and tensors should definitely support `.T` for transpose.