v0.4.6 - 2025-06-12

What's Changed

• Fixed an error when generating the parity plot for a specific range by @jcwang587 in #65
• Improved the CIFData caching mechanism by @jcwang587 in #76

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.4.5...v0.4.6

Changes

v0.4.5 - 2025-06-08

What's Changed

• Added a new section for training options in the documentation by @jcwang587 in #46
• Info displayed with logging by @jcwang587 in #47
• Added flexible size of caching for DataLoader by @jcwang587 in #52
• Simplified dataset splitting process for training, validation, and testing by @jcwang587 in #54
• Enhanced neighbor list loading with Cythonized get_neighbor_list by @jcwang587 in #60
• Changed the output format of parity plots from SVG to PNG by @jcwang587 in #63

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.4.4...v0.4.5

Changes

v0.4.4 - 2025-05-29

What's Changed

• Enhanced model inference efficiency by adopting torch.inference_mode by @jcwang587 in #38
• Add option to force training set inclusion while not preserving the train:valid:test ratio by @jcwang587 in #39
• Add print_checkpoint_info by @jcwang587 in #40
• Set shuffle=False for validation and test set by @jcwang587 in #42
• Remove scikit-learn dependency by @jcwang587 in #44
• Switch the build system from poetry to uv for package building and publishing by @jcwang587 in #45

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.4.3...v0.4.4

Changes

v0.4.3 - 2025-05-22

What's Changed

• Fix the error when inputting dataset without a split ratio by @jcwang587 in #36

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.4.2...v0.4.3

Changes

v0.4.2 - 2025-05-16

What's Changed

• Allow custom x‑ and y‑axis labels in parity plot by @jcwang587 in #33
• Sort output indices by structure ID when running predictions by @jcwang587 in #34

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.4.1...v0.4.2

Changes

v0.4.1 - 2025-04-29

What's Changed

• Add CLI scripts id_gen and atom_gen by @jcwang587 in #23
• Add documentation by @jcwang587 in #24
• Remove the requirement for id_prop.csv when running predictions on unknown datasets by @jcwang587 in #29

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.4.0...v0.4.1

Changes

v0.4.0 - 2025-04-09

What's Changed

• Add a function to deduplicate CIF files for cleaning the dataset by @jcwang587 in #8
• Set up pre-trained models in subpackage by @jcwang587 in #10
• Add some missing type hints to cgcnn_model.py by @Andrew-S-Rosen in #14
• Fix return value in p01_prediction.ipynb by @Andrew-S-Rosen in #11
• Sort imports and remove unused imports by @Andrew-S-Rosen in #12
• Use dict instead of Dict for type hinting by @Andrew-S-Rosen in #13
• Fix docstring for ConvLayer by @Andrew-S-Rosen in #17
• Add a cgcnn-pr cli script for prediction by @jcwang587 in #20
• Add a cgcnn-tr cli script for training from scratch by @jcwang587 in #22

New Contributors

• @Andrew-S-Rosen made their first contribution in #14

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.3.4...v0.4.0

Changes

v0.3.4 - 2025-03-26

What's Changed

• Add option to force training set inclusion while preserving the train:valid:test ratio by @jcwang587 in #7

Full Changelog: https://github.com/jcwang587/cgcnn2/compare/v0.3.2...v0.3.4

Changes

v0.3.2 - 2025-03-19

What's Changed

• Add an option to the fine-tuning script to generate a parity plot within a user-specified range

New Contributors

• @jcwang587 made their first contribution

Full Changelog: https://github.com/jcwang587/cgcnn2/commits/v0.3.2

Changes