Pyoperon in conda on Phoebe¶

Note

Note: This guide expects a clean shell environment. Do not run it inside Jupyter Notebooks or environments enriched with preloaded modules or dependencies.

Tested on: August 6, 2025

Build pyoperon and its dependencies¶

Clone pyoperon from the upstream repository¶

git clone https://github.com/heal-research/pyoperon.git
cd pyoperon

Activate conda and create the environment¶

Optional tweaks to `environment.yml`¶

Use a stable version: clangxx==19.1.7 instead of a cutting-edge version.
To prepare for openMPI later, you may add:

- ucx
- libpmix==5.0.8

📄 example environment.yaml can be found here.

module load Miniforge3
# this is alternative way to activate conda without modifying user configuration..:
source ${EBROOTMINIFORGE3}/etc/profile.d/conda.sh

conda env create -f environment.yml
conda activate pyoperon

Configure Clang as the compiler¶

export CC=${CONDA_PREFIX}/bin/clang
export CXX=${CONDA_PREFIX}/bin/clang++

Download and run the site-specific dependency script¶

wget https://gist.githubusercontent.com/jose-d/9db74a1283eba9fbadf73d2d029ad505/raw/4aaa3044909779675c57135f41cd6b35101522bb/dependencies.sh --output-document=./script/dependencies.sh
chmod +x ./script/dependencies.sh 
./script/dependencies.sh

✅ This step might take some time. Warnings are expected, but no critical errors should occur.

Install pyoperon¶

pip install .

Test the installation¶

Because the local directory shares a name with the module, avoid testing from within the project folder:

mkdir mytest
cd mytest
python

then test inside Python

>>> from pyoperon.sklearn import SymbolicRegressor
>>>

If no errors appear, the installation was successful!

Optional: install and run with OpenMPI¶

Install OpenMPI from source¶

🛠️ To ensure compatibility, you'll build OpenMPI and mpi4py within the Conda environment you created earlier.

Get script from here and run it within the conda environment created in previous steps. This will take several minutes:

wget https://gist.githubusercontent.com/jose-d/079c16c9bf767b243d1375c4267f4988/raw/d7b916da17e5e6d03c1b6d5c6eaef40895488eaf/install_ompi.sh --output-document=./install_ompi.sh
chmod +x ./install_ompi.sh
./install_ompi.sh

Install mpi4py from source¶

Once OpenMPI is built and available in your environment:

python -m pip install --no-binary=mpi4py mpi4py

This ensures mpi4py is compiled against your custom OpenMPI build.

Example batch job using the software built above¶

#!/bin/bash
#SBATCH --job-name=operon_testcase
#SBATCH --time=00:33:33
#SBATCH --partition=cpu
#SBATCH --ntasks=4
#SBATCH --cpus-per-task=64

echo "sbatch-INFO: start of job"
echo "sbatch-INFO: nodes: ${SLURM_JOB_NODELIST}"
echo "sbatch-INFO: system: ${SLURM_CLUSTER_NAME}"

module load Miniforge3
source ${EBROOTMINIFORGE3}/etc/profile.d/conda.sh
# here make sure you load the right enb
conda activate pyoperon_mpi
# the directory where you cloned pyoperon
cd $HOME/pyoperon
mpirun python ./hello_world.py



echo "sbatch-INFO: we're done"
date