Working repo for the PNNL MARS project "Probabilistic Model Discovery with Uncertainty Quantification"
With the .yml
file it suffices to say
$CONDA env create -f environment.yml
where $CONDA
is the conda
(Anaconda/Miniconda) binary. If the Anaconda binaries were added to $PATH
, then this is simply conda
.
If it hasn't been done before, it is necessary to set up the shell to activate and deactivate environments. For that, do
$CONDA init
$CONDA config --set auto_activate_base false
and restart the shell. The last line will prevent the base
environment from being loaded when the shell is launched (so that you have access from the shell to whatever Python you had first on your path, e.g., macOS's Python).
After restarting the shell, which conda
should definitely point to the conda
binary, and the remaining instructions assume that. If that's not the case, let David know!
Once the shell is set up, activating is done by
conda activate mars
which will add a (mars)
prefix to the prompt. After this, which python
and which jupyter
should point to the mars
environment's Python and Jupyter, respectively. You can then launch Jupyter the usual way: jupyter lab
.
To deactivate the environment, do
conda deactivate
After this, which python
should point to the whatever other Python you had on your path with first priority.
First check out sunode
and step onto the downloaded folder, e.g.:
cd ~/src
git clone https://github.com/aseyboldt/sunode.git
cd sunode
With conda
activated, and while stepping onto the sunode
source folder, create a new environment mars-sunode
:
conda create -n mars-sunode -q --yes -c conda-forge python=3.7 conda-build conda-verify coverage pytest hypothesis statsmodels
conda activate mars-sunode
conda-build -c conda-forge ./conda
conda install --yes -c $CONDA_PREFIX/conda-bld/ -c conda-forge sunode
conda install mkl-service
pip install pymc3 jupyterlab
After that you can deactivate and activate the mars-sunode
environment at your convenience.