Repo for data collection, discussion, etc for a MoSDeF reproducibility study.

We use conda to manage our software environment, so you'll need a conda installation--we recommend miniconda with mamba.

Once conda and mamba are installed, the mosdef-study38 environment can be created by running:

mamba env create -f environment.yml

And the environment can be activated by running:

conda activate mosdef-study38

The simulation engines used in this study require some additional setup. GROMACS, LAMMPS, HOOMD-blue, and Cassandra can be add to the environment by running:

mamba env update -f engines.yml

NOTE: For hoomd this installation may not work unless executed on a node with a GPU. To check GPU configuration run nvidia-smi. The hoomd gpu api version (python -c "import hoomd; print(hoomd.version.gpu_api_version)") should be the same as the cuda version.

GOMC can be installed according to its documentation here.

MCCCS can be installed according to its documentation here.

This project uses the Signac framework to manage its parameter space. Instructions for initializing and submitting/running the simulation and analysis workflows can be found in the project guide.

In order to share data between contributors, we use rclone to sync to a shared dropbox folder.

rclone can be installed without sudo permissions (as might be necessary, e.g., on a computing cluster) using conda:

conda install -c conda-forge rclone

Next rclone must be linked to dropbox. The next steps will use the remote name specified during the linking process. For the following examples we will use the name dropbox. If you forgot the name, you can use rclone listremotes.

Before syncing the workspace data, it is recommended to remove any empty job folders, as these will overwrite the data on dropbox. One way to do this is to remove the job folders of unused engines. We provide a bash script which can help(located in reproducibility_study/reproducibility_project/bin/clean_by_engine.sh). For example, to remove job folders for all engines EXCEPT hoomd, run:

bash bin/clean_by_engine.sh hoomd

Be careful when using this script as it could delete your data! Make sure to check which project (mosdef_reproducibility or lrc_shift_subproject) you are in using signac project. Before using it is advisable to make a backup and make sure that you spell/capitalize your engine the same way as in the statepoint.

Once your workspace contains only the data you want to sync, you can see how rclone would copy it to dropbox using the --dry-run flag:

rclone copy workspace dropbox:MoSDeF\ Repro\ Study\ Data/workspace --dry-run

The same command without --dry-run can be used to copy the data over.