DeepSim

Deep_Sim is a package to automatize the simulations (FEM and BEM) and to predict the quantities using Machine Learning and Deep Learning techniques.

---

                        *****    ******* *******  *****       *******  ** ***     ***
                        **  ***  **      **       **  **      **       ** ****   ****
                        **   *** *****   *****    *****       *******  ** ** ** ** **
                        **  ***  **      **       **               **  ** **  ***  **
                        *****    ******* *******  **          *******  ** **   *   **

---

                                    A Multipurpose Python automatization tool
                                    to run BEM and FEM simulations and apply
                                      Machine and Deep Learning techniques 

---

Contents

1. Prerequisites
2. Installation

Prerequisites

Deep_Sim is developed with python3.8. It is recommended to use virtual environments with python. Thus, the python3-venv package can be installed using the following command (On Debian systems):

sudo apt-get install python3-venv python3-dev

Installation

Clone the repo using recursive to get submodules as well:

git clone --recursive git@gitlab.com:compsim/codes/deep_sim.git

Create a new virtual environment using venv (for example in the home directory)

cd ~
mkdir opt
cd opt
python3 -m venv deep_sim_env

Activate the virtual environment:

source ~/opt/deep_sim_env/bin/activate

Go to the repository directory where you cloned to install the packages:

cd <deep_sim-directory>

Run the following command to install the default required packages. These packages are defined in the requirements.txt folder.

pip install -r requirements.txt

Finally setup the framework using python develop, which will install pydeep_sim and cubitpy packages.

python setup.py develop

List the installed packages

pip list --local

Working with submodules

DeepSim requires Cubitpy and Mirco to run simulations regarding FEM and BEM. To add those packages as submodules. If you have used the recursive option while cloning the repo, you can skip this part.

git submodule update --init --recursive

Working with CubitPy

To generate FEM models for Baci simulations, the CubitPy interface is used.

Add the CubitPy path to PYTHONPATH:

export PYTHONPATH=<path_to_cubitpy>:$PYTHONPATH

Cubitpy requires the pre_exodus executable and the cubit folder so set the paths for them:

export BACI_PRE_EXODUS=<path_to_pre-exodus>
export CUBIT=<path_to_cubit_directory>

Install the packages required for cubit:

pip install -r path_to_cubitpy/requirements.txt

Working with MIRCO

First, you have to build MIRCO by following the instructions on https://github.com/imcs-compsim/MIRCO. Then export the executable to bash:

export MIRCO=<path_to_mirco_executable>

Running the BACI code

To run Baci simulations, baci-release and post_drt_ensight executables should be exported as environment variables.

export BACI_RELEASE=<path_to_baci-release>

export BACI_POST_DRT_ENSIGHT=<path_to_post_drt_ensight>

Exporting results with Paraview

Since the simulation results of BACI have a special format, it is not possible to get the results without using Paraview or a post processing tool. Thus, it is possible to integrate Paraview to DeepSim. To do that

• export the path to PARAVIEW,

export PARAVIEWPATH=<path_to_paraview>

• add the site packages to PYTHONPATH,

export PYTHONPATH=$PYTHONPATH:$PARAVIEWPATH:/lib/python2.7/site-packages

• add the libraries of Paraview to LD_LIBRARY_PATH which looks for dynamic/shared libraries,

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${PARAVIEWPATH}/lib

Running tests

To check if the program works without any problem, tests are generated. At this point, Python offers the pytest interface. To run the tests:

pytest

If running all test scripts is not desired, then one can change the testpaths directory in the setup.cfg file for picking up specific tests.

Optional - But recommended

If you are a VSCode user and you don't want to export PYTHONPATH and all env variables all the time since you didn't export them permanently, i.e. .bashrc and more important you don't want to. There is a super easy way.

• Add the following lines to .vscode/settings.json: (Indeed, change the paths depending on your setup)

"terminal.integrated.env.linux": {
        "PARAVIEWPATH" : "/imcs/public/compsim/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit",
        "PYTHONPATH": "${env:PYTHONPATH}:/home/a11btasa/deep_sim/cubitpy/:/imcs/public/compsim/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit/lib/python3.8/site-packages",
        "LD_LIBRARY_PATH" : "${env:LD_LIBRARY_PATH}:/imcs/public/compsim/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit/lib",
        "BACI_RELEASE" : "/home/a11btasa/git_repos/baci/baci_build/baci-release",
        "BACI_POST_DRT_ENSIGHT" : "/home/a11btasa/git_repos/baci/baci_build/post_drt_ensight",
        "CUBIT": "/imcs/public/compsim/opt/cubit-13.2",
        "BACI_PRE_EXODUS": "/home/a11btasa/git_repos/baci/baci_build/pre_exodus",
        "BEM" : "/home/a11btasa/git_repos/bem/bem",
        "MIRCO": "/home/a11btasa/deep_sim/MIRCO/mirco"
    },
"python.envFile": "${workspaceFolder}/debug.env",

• Create the debug.env file where you clone the repo and copy the following lines there: (Indeed, again change the paths)

PARAVIEWPATH=/imcs/public/compsim/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit
PYTHONPATH=${PYTHONPATH}:/home/a11btasa/deep_sim/cubitpy/:/imcs/public/compsim/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit/lib/python3.8/site-packages
LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/imcs/public/compsim/opt/ParaView-5.9.1-MPI-Linux-Python3.8-64bit/lib
BACI_RELEASE=/home/a11btasa/git_repos/baci/baci_build/baci-release
BACI_POST_DRT_ENSIGHT=/home/a11btasa/git_repos/baci/baci_build/post_drt_ensight
CUBIT=/imcs/public/compsim/opt/cubit-13.2
BACI_PRE_EXODUS=/home/a11btasa/git_repos/baci/baci_build/pre_exodus
BEM=/home/a11btasa/git_repos/bem/bem
MIRCO="/home/a11btasa/deep_sim/MIRCO/mirco"

The settings.json will set the variables for the integrated shell, while debug.env will set the variable but for the debugging menu and they must be done seperately.

Note: If you are not a linux user, eg. windows, then set terminal.integrated.env.windows.

---

References:

• Environment variables in VSCODE

Docker installation (on test)

# From your Deep_Sim repo root (with submodules already pulled or not—both OK)
docker build -t deep_sim:dev .

# Quick sanity run (uses the CMD)
docker run --rm deep_sim:dev