Computational Graph States for Variational Monte Carlo

Computational framework that implements Computational Graph States (CGS) and Supervised Wavefunction Optimization (SWO).

Installation:

This project uses Bazel, but can be also executed directly using python3.

The following libraries are required to run the code:

• TensorFlow
• Sonnet
• absl-py

Can be obtained using pip.

Running:

cd cgs_vmc
bazel run :run_training -- --checkpoint_dir=PATH_TO_CHECKPOINT_DIRECTORY

To change optimization methods, variational ansatz etc. Provide corresponding flags to the run_training call.

Current implementation does not work with $log$ of the wavefunctions and therefore might experience numerical instabilities for large system sizes. This can be tuned out by appropriate normalization, but will be removed in future releases.

For suggestions and comments reach out to kochkov92