Haoyin Xu

@sdruskat Thanks! Please move the issue as you see fit.

By the way, I used `CFF 1.1.0` because the `cffinit` [website](https://citation-file-format.github.io/cff-initializer-javascript/) only supports that version.

Adding a `CITATION.cff` could also specify authors & orcids.

@srecupe2 are you still working on #213?

Example: https://github.com/neurodata/ProgLearn/pull/547/commits/bbf358c7738529d36ed5d232cc5120eb3451db75

@NogaMudrik cool. You can do so after hyperparameter tuning and tabnet.

Was this commented out when you run `create_train_loader`? ```python # for i in train_idxs: # dataset.targets[i] = np.where(classes == dataset.targets[i])[0][0] ```

You need to use deep copies of the dataset on that, too. There's no point of deep copies if you don't save the original. Then the copy becomes "original."

You have to create a deep copy each time you use those functions.

Create deep copies inside the functions