CounterNet: End-to-End Training of Counterfactual Aware Predictions

This is the official repository of the paper CounterNet: End-to-End Training of Counterfactual Aware Predictions. The purpose of the repository is only for research and reproduction of the paper's results. The audience should not expect to use the code directly in the deployed environemnt.

The repository is built based on nbdev. I highly recommend you to check out nbdev if you enjoy writing code with Jupyter Notebook as I do. Further, this project primarily leverages Pytorch and Pytorch Lightning for implementations of deep learning models. To install all the dependencies, you should run:

pip install -e .

Note:

• pip install will only install cpu-version of pytorch. If you want to use GPU-version of pytorch, please follow pytorch's official instruction.
• As Pytorch Lightning's API changes rapidly, it is not guaranteed that the code is compatible with other versions of Lightning (except the version that specified settings.ini).

Useful commands for nbdev

Build nbs to module

nbdev_build_lib

Update nbs from module

nbdev_update_lib

clean notebooks

nbdev_clean_nbs

Test all modules

nbdev_test_nbs

Start tensorboard

tensorboard --logdir log --bind_all