This is a gaussian emulator and Baysiean analysis with PyMC
- Install miniforge for python3 by following the instructions on https://github.com/conda-forge/miniforge.
- Update and activate the base environment:
$> conda update --all
$> conda activate
- Set the channel priority to flexible (required before using the environment yml files, otherwise the solver may fail or produce an unusable env):
$> conda config --set channel_priority flexible
-
Go to /path/to/Bayesian-package and create the virtual environment.
On Linux/Windows:
$> conda env create --name <env_name> --file=environmentCrossPlatform.yml
On Mac, environmentCrossPlatform.yml does NOT work — use environmentMac.yml instead, and prefix the command with CONDA_SUBDIR=osx-64 so the correct platform packages are selected (without it, packages will fail to install):
$> CONDA_SUBDIR=osx-64 conda env create --name <env_name> --file=environmentMac.yml
- Activate the environment:
$> conda activate <env_name>
-
Start the program.
On Linux/Windows:
$> python -m GUI.GUIController.GUIControllerMP <number-of-cores>
On Mac, you must use pythonw instead of python (the regular python will not work for the GUI):
$> pythonw -m GUI.GUIController.GUIControllerMP <number-of-cores>
IF the script refuses to use more than one core despite > 1, try to restart the terminal after step 4. Sometimes the installation of conda messes up the CPU affinity. Restarting after installation usually fixes it.
If not, use
$> OPENBLAS_MAIN_FREE=1 python -m GUI.GUIController.GUIControllerMP <number-of-cores>
(or pythonw instead of python on Mac).
TBD