eth-sri/abstraqt

Abstraqt

Dependencies

We provide a docker image providing all dependencies for Abstraqt. To create the docker image and enter its container, run (you may need sudo):

make -C docker run

Alternatively, to develop Abstraqt, install conda and the dependencies in the Dockerfile.

Installation

Inside the docker container, run:

time bash ./install-pip.sh  # install Abstraqt
rm -rf QuiZX/quizx && git clone https://github.com/Quantomatic/quizx.git QuiZX/quizx && cd QuiZX/quizx && git checkout 81e9e63 && cd - && make -C QuiZX/wrapper  # build QuiZX
cd baseline_yp21/v8; bash ./clean.sh; javac Main.java; cd -  # build YP21

This takes around half a day, as it includes running extensive tests and some pre-computation.

Usage

from abstraqt.stabilizer.abstract_stabilizer_plus import AbstractStabilizerPlus


state = AbstractStabilizerPlus.zero_state(2)

state = state.conjugate('X', 0)
state = state.conjugate('CNOT', 0, 1)
state = state.conjugate('T', 0)
state = state.compress(1)
state = state.measure('Z', 0)  # project to |0⟩
probability_range = state.get_trace()

print(probability_range)

Evaluation

To reproduce the results from our evaluation of Abstraqt, simply run

/usr/bin/time -v python abstraqt/experiments/run_experiments.py --max-processes 1

This takes multiple days, even when using 20 processes (see flag above), as it includes running multiple slow baselines. Using 20 processes is only an option if enough cores (>20) and enough memory (~100GB) is available.

Memory Usage

To determine the memory usage of Abstraqt, run:

/usr/bin/time -v python abstraqt/experiments/run_experiments.py --tool abstraqt

Generate Benchmarks

To regenerate a fresh benchmark from scratch:

# optional: cleanup old benchmarks
cd abstraqt/experiments/circuits; rm *.qasm *.q *.qc *.csv; git checkout *.qasm *.q *.qc *.csv; cd -; git status
# generate
python abstraqt/experiments/generate_circuits.py --repeat 10