Study Case: Reinforcement Learning for Fluids¶

Dependencies¶

CubismUP-2D has the following prerequisite libraries:

MPI, with the $MPICXX enviroment variable defined.
GSL, with the $GSL_ROOT environment variable defined.
HDF5, with the $HDF5_ROOT environment variable defined.

On Piz Daint:

module load module load daint-gpu GSL cray-hdf5-parallel cray-fftw cray-python
export GSL_ROOT=/apps/dom/UES/jenkins/7.0.UP02/gpu/easybuild/software/GSL/2.5-CrayGNU-20.11
export MPICXX=CC
export CC=cc
export CXX=CC
export OMP_NUM_THREADS=12

On Panda/Falcon:

module load gnu mpich python hdf5
export GSL_ROOT=/usr

Setup¶

Install CubismUP-2D by running:

./install_cup.sh

After installing the flow solver you can go to either of the provided examples and compile them using

cd EXAMPLE
make -j

Running¶

The examples can be run using the provided scripts in the subdirectories. See the README there.