hopfield-tracking

Some experiments with Hopfield networks for tracking

Local development

Prepare

conda env create
conda activate hopfield-tracking
nb-clean add-filter --remove-empty-cells

Experiment

• easy generated tracks: jupyter lab main.ipynb
• noisy real data: jupyter lab real-events.ipynb

Test locally

pytest

Jhub on hybrilit

First setup

wget -c https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
## !! install to /tmp/$USER/miniconda3 !!
bash Miniconda3-latest-Linux-x86_64.sh
/tmp/$USER/miniconda3/bin/conda env create -f env-jhub.yml
/tmp/$USER/miniconda3/envs/hopfield-tracking/bin/nb-clean add-filter --remove-empty-cells
git config --local --add filter.nb-clean.clean "/tmp/$USER/miniconda3/envs/hopfield-tracking/bin/nb-clean clean --remove-empty-cells"
/tmp/$USER/miniconda3/envs/hopfield-tracking/bin/python -m ipykernel install --user --name=hopfield-tracking
tar -cf hopfield-tracking.env.tar.lz4 -I lz4 -C /tmp/$USER/miniconda3/envs hopfield-tracking
tar -cf miniconda3.tar.lz4 -I lz4 -C /tmp/$USER/ miniconda3

Test on hydra client node

sh hydra-env.sh
/tmp/$USER/miniconda3/envs/hopfield-tracking/bin/pytest

Experiment interactively

• https://jhub.jinr.ru
• main.ipynb
• real-events.ipynb

Run in Slurm

• https://jlabhpc.jinr.ru
• sbatch.ipynb