v0.0.7

tabben v0.0.7

New Features

• added the CIFAR10 dataset as a standardized dataset (i.e. can be accessed directly using OpenTabularDataset)
• added a DatasetCollection class for bulk processing of datasets
• multiple splits can be passed into the OpenTabularDataset constructor, e.g. OpenTabularDataset(dir, name, split=['train', 'valid'])

Breaking Changes

• removed the TabularCIFAR10Dataset as it is no longer needed
• categorical columns for all datasets start counting at 0
• these OpenTabularDataset constructor parameters are now keyword only: download, transform, target_transform

Non-Breaking Changes

• many datasets have additional "extras" (every current dataset has extras) available, such as
  • training data profiles (and full-data profiles that should not be used for model selection)
  • bibtex, licenses when available
• as a result of the many changes to several datasets, the version for all datasets has been incremented
  • version numbers are now 3 element integer arrays

Bugfixes

Non-Code Updates

• the documentation website is now generated with sphinx, and has API reference info as well
• dataset pages for amazon and rossman datasets added to docs website
• there is now also a Julia package Tabben.jl for loading datasets and evaluating models (not at feature parity yet)