Instructions to reproduce this benchmark

(a) Overhead of individual quantities

1. Run experiment: python run_individual.py. Alternatively, extract original data: unzip results_individual.zip into the results directory.
2. Create plots and move them to the tex directory: python plot_individual.py. Plots can be found in output/fig_individual.
3. Clean up or start over: bash clean_individual.sh

(b) Overhead of configurations at different tracking rates

1. Run experiment: python run_grid.py and python run_grid_imagenet.py. Alternatively, extract original data: unzip results_grid.zip into the results directory.
2. Create plots and move them to the tex directory: python plot_grid.py. Plots can be found in output/fig_grid.
3. Clean up or start over: bash clean_grid.sh