0.4.0 tasks (gnn update)

[wolearyc]

For now, the model itself is only tested on TiO2. Future versions will test additional systems, including enhancements as necessary.

New Features

• Graph neural network polarizability model
• Build system with fixed pytorch installation (check python version) as well as special handling for pytorch extensions, like pytorch_geometric, pytorch_scatter, and pytorch_sparse
• read_structure_and_polarizability functions for vasprun.xml files.
• [torch] install options 1148a0e, 8aa0287
• Training module

API Changes

• UsageError --> UserError
• File renames: io_utils, spectrum_utils, structure_utils --> utils
• Rename globals --> constants
• polarizability --> pmodel

Enhancements and Bugfixes

• add robot.txt for docs 9f78873
• Exception is raised when filepaths is empty 73fcd16.
• Clean up imports - should we make all submodules available?
• close(...).all() --> allclose(...)

Documentation

• Standardize tuple documentation, including mentioning the size of the tuple.
• Installation guide
• Update citation
• Remove '1D', '2D', '3D' etc.
• Machine learning tutorial