ACTS-Exatrkx Integration

Table of Contents

• General Info
• Cori Setup
  • Starting CVMFS Container
  • Starting Python VirtualEnv
• Install
  • Installing Exatrkx-Iml2020
  • Compiling ACTS
• Getting Started
  • Simulating Basic Dataset
  • Run Track Reconstruction Pipeline
• Documentation
  • Track Reconstruction Pipeline
  • InferTracks
  • GraphNeuralNetwork

General Info

The goal of this project is to integrate the Exatrkx inference pipeline into ACTS. This is a working repository and is still under development.

Cori Setup

This section gives step by step instructions for installing dependencies for ACTS and Exatrkx on Cori.

Starting CVMFS Container

First time setup - only run the following commands once:

rm -rf ~/.alrb
chmod g+s $HOME
setfacl -d -m u:nobody:rx $HOME
setfacl -m u:nobody:rx .

Every time you log in:

source /global/project/projectdirs/atlas/scripts/setupATLAS.sh
setupATLAS -c centos7+batch

You should now be running inside of a container. To load the ACTS dependencies, run the following:

source /cvmfs/sft.cern.ch/lcg/views/LCG_97apython3/x86_64-centos7-gcc9-opt/setup.sh
export PATH=/cvmfs/sft.cern.ch/lcg/releases/CMake/3.17.3-75516/x86_64-centos7-gcc9-opt/bin:$PATH

You should now have a working container with access to all neccessary ACTS dependencies.

Starting Python VirtualEnv

We now must create a Python virtual environment to give us the freedom to install Exatrkx's dependencies.

python -m venv inference-pipeline-venv

Then to load the virtual environment:

source inference-pipeline-venv/bin/activate

When the container from the previous is initialized, it makes edits to Python environment variables which will become a problem later. These are removed with the following commands:

unset PYTHONPATH
export LD_LIBRARY_PATH=/cvmfs/sft.cern.ch/lcg/releases/java/8u222-884d8/x86_64-centos7-gcc9-opt/jre/lib/amd64: \
                       /cvmfs/sft.cern.ch/lcg/views/LCG_97apython3/x86_64-centos7-gcc9-opt/lib64: \
                       /cvmfs/sft.cern.ch/lcg/views/LCG_97apython3/x86_64-centos7-gcc9-opt/lib: \
                       /cvmfs/sft.cern.ch/lcg/releases/gcc/9.2.0-afc57/x86_64-centos7/lib: \
                       /cvmfs/sft.cern.ch/lcg/releases/gcc/9.2.0-afc57/x86_64-centos7/lib64: \
                       /cvmfs/sft.cern.ch/lcg/releases/binutils/2.30-e5b21/x86_64-centos7/lib: \
                       /cvmfs/sft.cern.ch/lcg/releases/R/3.6.3-ca0ad/x86_64-centos7-gcc9-opt/lib64/R/library/readr/rcon

You should now have a container loaded with ACTS dependencies and a Python virtual environment ready for Exatrkx's dependencies.

Install

This section guides the user through the installation of ACTS and Exatrkx's inference pipeline. It is assumed the user already has ACTS's dependencies installed.

Installing Exatrkx Inference Pipeline

Go to https://github.com/sbconlon/exatrkx-iml2020 and follow the setup steps. Now add inference_fn.py to the PYTHONPATH environment variable.

export PYTHONPATH=/path/to/exatrkx-iml2020/notebooks:$PYTHONPATH

Finally, you should be able to run the following command from any directory without error:

python -c "from inference_fn import gnn_track_finding"

Compiling ACTS

From your build directory:

cmake -DACTS_BUILD_EXAMPLES=ON -DACTS_BUILD_UNITTESTS=ON -DACTS_BUILD_EXAMPLES_EXATRKX=ON -DACTS_BUILD_PLUGIN_EXATRKX=ON /path/to/acts-source
make -j30

Or if you are running inside a Python virtual environment:

cmake -DACTS_BUILD_EXAMPLES=ON -DACTS_BUILD_UNITTESTS=ON -DACTS_BUILD_EXAMPLES_EXATRKX=ON -DACTS_BUILD_PLUGIN_EXATRKX=ON -DPython3_FIND_VIRTUALENV=ONLY /path/to/acts-source

You are now ready to reconstruct tracks.

Getting Started

This section guides a user through running the ACTS+Exatrkx example reconstruction pipeline on a basic dataset.

Simulating a Basic Dataset

Simulate single muon events inside the generic, TrackML-like detector. From the bin directory:

./ActsExampleFatrasGeneric \
    --output-dir=/path/to/data/sim-generic/single-muon \
    --output-csv \
    --events=10 \
    --bf-value=0 0 2

Run Track Reconstruction Pipeline

Run the ACTS+Exatrkx example reconstruction pipeline.

./ActsExampleGNNTracksGeneric \
    --input-dir=/path/to/data/sim-generic/single-muon \
    --output-dir=/path/to/output \
    --digi-smear \
    --digi-config-file=/path/to/acts-source/Examples/Algorithms/Digitization/share/default-smearing-config-generic.json \
    --ml-module-name=inference_fn \
    --ml-function-name=gnn_track_finding \
    -N 1 -j 1

Performance plots are written to given output directory.

Documentation

Track Reconstruction Pipeline

InferTracks

GraphNeuralNetwork Class