<!-- These are examples of badges you might also want to add to your README. Update the URLs accordingly. -->

⚠️ Manuscript and results using a generative model coming soon ⚠️

xtal2png

Encode/decode a crystal structure to/from a grayscale PNG image for direct use with image-based machine learning models such as Google's Imagen.

The latest advances in machine learning are often in natural language such as with LSTMs and transformers or image processing such as with GANs, VAEs, and guided diffusion models. Encoding/decoding crystal structures via grayscale PNG images is akin to making/reading a QR code for crystal structures. This allows you, as a materials informatics practitioner, to get streamlined results for new state-of-the-art image-based machine learning models applied to crystal structure. Let's take Google's text-to-image diffusion model, Imagen (unofficial), which can also be used as an image-to-image model. Rather than dig into the code spending hours, days, or weeks modifying, debugging, and playing GitHub phone tag with the developers before you can (maybe) get preliminary results, xtal2png lets you get those results using the default instructions on the repository.

After getting preliminary results, you get to decide whether it's worth it to you to take on the higher-cost/higher-expertise task of modifying the codebase and using a more customized approach. Or, you can stick with the results of xtal2png. It's up to you!

Getting Started

Installation

bash conda create -n xtal2png -c conda-forge xtal2png m3gnet conda activate xtal2png

NOTE: m3gnet is an optional dependency that performs surrogate DFT relaxation.

Example

Here, we use the top-level XtalConverter class with and without optional relaxation via m3gnet.

```python

example_structures is a list of pymatgen.core.structure.Structure objects

from xtal2png import XtalConverter, example_structures

xc = XtalConverter(relaxondecode=False) data = xc.xtal2png(examplestructures, show=True, save=True) decodedstructures = xc.png2xtal(data, save=False) len(decoded_structures) 2

xc = XtalConverter(relaxondecode=True) data = xc.xtal2png(examplestructures, show=True, save=True) relaxeddecodedstructures = xc.png2xtal(data, save=False) len(relaxeddecoded_structures) 2

```

Output

python print(example_structures[0], decoded_structures[0], relaxed_decoded_structures[0])

The before and after structures match within an expected tolerance; note the round-off error due to encoding numerical data as RGB images which has a coarse resolution of approximately 1/255 = 0.00392. Note also that the decoded version lacks charge states. The QR-code-like intermediate PNG image is also provided in original size and a scaled version for a better viewing experience: | 64x64 pixels | Scaled for Better Viewing (tool credit) | Legend | | ----------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------- | | | | |

Additional examples can be found in the docs.

Limitations and Design Considerations

There are some limitations and design considerations for xtal2png. Here, we cover round-off error, image dimensions, contextual features, and customization.

Round-off

While the round-off error is a necessary evil for encoding to a PNG file format, the unrounded NumPy arrays can be used directly instead if supported by the image model of interest via structures_to_arrays and arrays_to_structures.

Image dimensions

We choose a $64\times64$ representation by default which supports up to 52 sites within a unit cell. The maximum number of sites max_sites can be adjusted which changes the size of the representation. A square representation is used for greater compatibility with the common limitation of image-based models supporting only square image arrays. The choice of the default sidelength as a base-2 number (i.e. $2^6$) reflects common conventions of low-resolution images for image-based machine learning tasks.

Contextual features

While the distance matrix does not directly contribute to the reconstruction in the current implementation of xtal2png, it serves a number of purposes. First, similar to the unit cell volume and space group information, it can provide additional guidance to the algorithm. A corresponding example would be the role of background vs. foreground in classification of wolves vs. huskies; oftentimes classification algorithms will pay attention to the background (such as presence of snow) in predicting the animal class. Likewise, providing contextual information such as volume, space group, and a distance matrix is additional information that can help the models to capture the essence of particular crystal structures. In a future implementation, we plan to reconstruct Euclidean coordinates from the distance matrices and homogenize (e.g. via weighted averaging) the explicit fractional coordinates with the reconstructed coordinates.

Customization

See the docs for the full list of customizable parameters that XtalConverter takes.

Installation

PyPI (pip) installation

Create and activate a new conda environment named xtal2png (-n) with python==3.9.* or your preferred Python version, then install xtal2png via pip.

bash conda create -n xtal2png python==3.9.* conda activate xtal2png pip install xtal2png

Editable installation

In order to set up the necessary environment:

1. clone and enter the repository via:

bash git clone https://github.com/sparks-baird/xtal2png.git cd xtal2png

1. create and activate a new conda environment (optional, but recommended)

bash conda env create --name xtal2png python==3.9.* conda activate xtal2png

1. perform an editable (-e) installation in the current directory (.):

bash pip install -e .

NOTE: Some changes, e.g. in setup.cfg, might require you to run pip install -e . again.

Optional and needed only once after git clone:

1. install several pre-commit git hooks with:

bash pre-commit install # You might also want to run `pre-commit autoupdate`

and checkout the configuration under .pre-commit-config.yaml. The -n, --no-verify flag of git commit can be used to deactivate pre-commit hooks temporarily.

1. install nbstripout git hooks to remove the output cells of committed notebooks with:

bash nbstripout --install --attributes notebooks/.gitattributes

This is useful to avoid large diffs due to plots in your notebooks. A simple nbstripout --uninstall will revert these changes.

Then take a look into the scripts and notebooks folders.

Command Line Interface (CLI)

Make sure to install the package first per the installation instructions above. Here is how to access the help for the CLI and a few examples to get you started.

Help

You can see the usage information of the xtal2png CLI script via:

bash xtal2png --help

```bash Usage: xtal2png [OPTIONS]

xtal2png command line interface.

Options: --version Show version. -p, --path PATH Crystallographic information file (CIF) filepath (extension must be .cif or .CIF) or path to directory containing .cif files or processed PNG filepath or path to directory containing processed .png files (extension must be .png or .PNG). Assumes CIFs if --encode flag is used. Assumes PNGs if --decode flag is used. -s, --save-dir PATH Encode CIF files as PNG images. --encode Encode CIF files as PNG images. --decode Decode PNG images to CIF files. -v, --verbose TEXT Set loglevel to INFO. -vv, --very-verbose TEXT Set loglevel to INFO. --help Show this message and exit. ```

Examples

To encode a single CIF file located at src/xtal2png/utils/Zn2B2PbO6.cif as a PNG and save the PNG to the tmp directory:

bash xtal2png --encode --path src/xtal2png/utils/Zn2B2PbO6.cif --save-dir tmp

To encode all CIF files contained in the src/xtal2png/utils directory as a PNG and save corresponding PNGs to the tmp directory:

bash xtal2png --encode --path src/xtal2png/utils --save-dir tmp

To decode a single structure-encoded PNG file located at data/preprocessed/Zn8B8Pb4O24,volume=623,uid=b62a.png as a CIF file and save the CIF file to the tmp directory:

bash xtal2png --decode --path data/preprocessed/Zn8B8Pb4O24,volume=623,uid=b62a.png --save-dir tmp

To decode all structure-encoded PNG file contained in the data/preprocessed directory as CIFs and save the CIFs to the tmp directory:

bash xtal2png --decode --path data/preprocessed --save-dir tmp

Note that the save directory (e.g. tmp) including any parents (e.g. ab/cd/tmp) will be created automatically if the directory does not already exist.

Project Organization

├── AUTHORS.md <- List of developers and maintainers. ├── CHANGELOG.md <- Changelog to keep track of new features and fixes. ├── CONTRIBUTING.md <- Guidelines for contributing to this project. ├── Dockerfile <- Build a docker container with `docker build .`. ├── LICENSE.txt <- License as chosen on the command-line. ├── README.md <- The top-level README for developers. ├── configs <- Directory for configurations of model & application. ├── data │ ├── external <- Data from third party sources. │ ├── interim <- Intermediate data that has been transformed. │ ├── preprocessed <- The final, canonical data sets for modeling. │ └── raw <- The original, immutable data dump. ├── docs <- Directory for Sphinx documentation in rst or md. ├── environment.yml <- The conda environment file for reproducibility. ├── models <- Trained and serialized models, model predictions, │ or model summaries. ├── notebooks <- Jupyter notebooks. Naming convention is a number (for │ ordering), the creator's initials and a description, │ e.g. `1.0-fw-initial-data-exploration`. ├── pyproject.toml <- Build configuration. Don't change! Use `pip install -e .` │ to install for development or to build `tox -e build`. ├── references <- Data dictionaries, manuals, and all other materials. ├── reports <- Generated analysis as HTML, PDF, LaTeX, etc. │ └── figures <- Generated plots and figures for reports. ├── scripts <- Analysis and production scripts which import the │ actual PYTHON_PKG, e.g. train_model. ├── setup.cfg <- Declarative configuration of your project. ├── setup.py <- [DEPRECATED] Use `python setup.py develop` to install for │ development or `python setup.py bdist_wheel` to build. ├── src │ └── xtal2png <- Actual Python package where the main functionality goes. ├── tests <- Unit tests which can be run with `pytest`. ├── .coveragerc <- Configuration for coverage reports of unit tests. ├── .isort.cfg <- Configuration for git hook that sorts imports. └── .pre-commit-config.yaml <- Configuration of pre-commit git hooks.

Note on PyScaffold

This project has been set up using PyScaffold 4.2.1 and the dsproject extension 0.7.1.

To create the same starting point for this repository, as of 2022-06-01 on Windows you will need the development versions of PyScaffold and extensions, however this will not be necessary once certain bugfixes have been introduced in the next stable releases:

bash pip install git+https://github.com/pyscaffold/pyscaffold.git git+https://github.com/pyscaffold/pyscaffoldext-dsproject.git git+https://github.com/pyscaffold/pyscaffoldext-markdown.git

The following pyscaffold command creates a starting point for this repository:

bash putup xtal2png --github-actions --markdown --dsproj

Alternatively, you can edit a file interactively and update and uncomment relevant lines, which saves some of the additional setup:

bash putup --interactive xtal2png

Attributions

• @michaeldalverson for iterating through various representations during extensive work with crystal GANs. The base representation for xtal2png (see #output) closely follows a recent iteration (2022-06-13), taking the first layer ($1\times64\times64$) of the $4\times64\times64$ representation and replacing a buffer column/row of zeros with unit cell volume.