This the the source for the numerical experiments found in the arxiv paper Iterative Surrogate Model Optimization (ISMO): An active learning algorithm for PDE constrained optimization with deep neural networks arXiv:2008.05730
Clone with
git clone --recursive git@github.com:kjetil-lye/ismo_airfoil.git
If you already cloned without the recursive option, do (from ismo_airfoil):
git submodule update --init --recursive
To make sure one has all the python packages required (and that one does not mess up ones python directory), one can use virtualenv. First install it (for python3) :
pip3 install --user virtualenv
or you can leave out the --user option if you want it to be available for all users.
Then create a new virutal environment (see the documentation for what is going on):
virtualenv3 .venv
activate the environment
source .venv/bin/activate
Then, only for the first time, install the needed packages (after doing souruce .venv/bin/activate):
pip install -r requirements.txt
Now you can run the commands below. To leave the virtual enviroment, use
deactivate
which will give you back the ordinary shell.
In general, to use virutalenv from a new terminal window once you have the .venv setup, you do
cd <path to ismo_airfoil>
source .venv/bin/activate
# Now you are in the virtual environemnt (promp should have a (.venv) in it)
cd airfoil_chain
bash run_in_bash.sh
# do some more
deactivate
All the commands should be run in the subfolder airfoil_chain.
To run outside of Euler/LSF-clusters
bash run_in_bash.sh
To just display the commands that would be run without running, do
bash run_in_bash.sh --dry_run
To run on Euler/LSF do
bash run_on_euler.sh
and again to only display the bsub commands that would be run do
bash run_on_euler.sh --dry_run