Installation

Using virtualenv

Pre-requisities

Python >= 3.9
NVIDIA GPU with CUDA installed
dist-utils: sudo apt install python3-distutils
virtualenv: pip install virtualenv

Obtaining the package

Change to a directory where you want to place the TPCwithDNN package. In there, type

git clone https://github.com/AliceO2Group/TPCwithDNN.git

and you will find a new directory TPCwithDNN.

RootInteractive

Note: Parts of RootInteractive depend on AliRoot, so you need to build and source AliRoot to use some of the features. These features are used in the processing of the validation data to create nd histograms and pdf maps.

python3 -m pip install git+https://github.com/miranov25/RootInteractive

Prepare the virtual environment

To keep the python environment clean, we will use a virtual environment for the software setup. Dealing with that is handled by TPCwithDNN/load.sh and there are 3 scenarios when you type:

source load.sh # This you can actually source from anywhere

If the environment has not been created yet, sourcing will create and enter it.
Being inside the environment, sourcing it again will leave the environment.
Outside of the environment, source load.sh --recreate will delete and re-install it.

At the beginning of load.sh you should check and adjust the paths to your CUDA and AliRoot installations.

From inside the virtual environment, change to TPCwithDNN/ where you can find setup.py. To install the package with all its dependencies, run

pip install -e .

Using singularity containers

Instructions on how to build the singularity container and the required software are collected in the alice-tpc-offline code repository in the directory JIRA/ATO-500/machineLearning. The following readme gives a compact summary: https://gitlab.cern.ch/alice-tpc-offline/alice-tpc-notes/-/blob/master/JIRA/ATO-500/machineLearning/READMECompact.md.