paces:

Parallelized Application of Co-Evolving Subspaces, a method for computing quantum dynamics on GPUs

The main branch will be updated with new features to support a wider variety of models. If you are interested in the single-exciton Holstein version that was used to create the initial data shown in the publication [link to be inserted here], please see the holstein_only branch instead.

Dependencies:

This code requires SciPy and CuPy and its dependencies, chiefly Python 3, NumPy, CUDA. It is known to work with the following combination of versions, but will most likely work with other versions as well:

• CuPy 11.2
• CUDA 11.8 with cuDNN 8.4.0.27, cuTENSOR 1.5.0.3, NCCL 2.14.3
• Python 3.10.8
• NumPy 1.24.1
• SciPy 1.10.0

Initial (one-time) setup:

After having ensured that the dependencies are met: Copy device_config_template.py to device_config.py and make necessary changes according to your system setup (number of GPUs, memory). device_config.py and any changes made to it will not be synced via git.

Per-calculation setup:

Before starting a calculation, first create a suitable subdirectory structure to store the data in.

As an example, we assume you are located in the directory containing this README.md and say we want to store our files in ../parent_dir/main_calc_dir. Then the steps are:

1. Ensure that ../parent_dir already exists.
2. Create the actual calculation directory itself including necessary subdirectories using bash prepare_dir.sh ../parent_dir/main_calc_dir.
3. Set the parameters of the calculation as desired in the short wrapper file main.py.
4. Run the calculation using python3 main.py.

By default, the calculation directory specified within main.py is given as a relative path to main.py itself, not to the location from which it is called. To change this behavior, remove the os.chdir call from the start of the file.