Conda environment creation using an Aqueduct base environment

conda-optimized.sh

# Create a base Conda environment with Python 3.9 and the aqueduct-ml package.
# These steps sit outside of the critical path of running a function.
conda create -n aqueduct_py39 python==3.9
conda run -n aqueduct_py39 pip install aqueduct-ml

# Since the Python 3.9 runtime was already downloaded and cached when building
# the base image, this step now only takes 5 seconds.
conda create -n my_workflow_env python==3.9

# This command adds the site package path of aqueduct_py39 to the PYTHONPATH of
# my_workflow_env so that it can access packages installed in aqueduct_py39.
# Essentially, this command allows my_workflow_env to inherit aqueduct_py39.
# This step takes 1.1 seconds.
conda develop -n my_workflow_env <path_to_aqueduct_py39>/lib/python3.9/site-packages

# Now we just need to install the scikit-learn package.
conda run -n my_workflow_env pip install scikit-learn