Running your first quantum computing job on Helmi through LUMI

If you've applied for a project, been accepted, setup your ssh keys and gained access to LUMI, then the next step is to run your first quantum computing job on a real quantum computer! This is a guide for exactly how to do that. The only thing you need to know is your project number!

Configuring the environment

The first step after you have logged into LUMI (via ssh lumi on your terminal) is to configure the environment. The base environment when first logging into LUMI does not provide the necessary tools to submit quantum jobs, therefore a quantum software stack has been created which sets up the correct python virtual environments and the correct environment variables. This is accessed through the LMOD system on LUMI using modules.

To use the quantum software stack you first need to tell LMOD where to search for modules.

module use /appl/local/quantum/modulefiles

You can then see the list of available modules with module avail. The quantum modules should be at the top! In this walkthrough Qiskit will be used, therefore the next step is to load the module into our current environment with

module load helmi_qiskit

Creating your first quantum program

The next step is to create your quantum circuit! Here a simple bell state will be created between two qubits, demonstrating entanglement between them! For this we will be using Qiskit but the steps are very similar for Cirq. It is good practice to work in your projects scratch directory, which you can navigate to with cd /scratch/project_xxx, inserting your project number.

Tip!

You can quickly see your LUMI workspaces with module load lumi-workspaces and lumi-workspaces

Let us first create our python file with nano my-first-quantum-job.py. Here we use nano but if you are comfortable you can also use vim or emacs. This will bring up the nano text editor, the useful commands are at the bottom, to save and exit CTRL-X + Y.

Importing the libraries

First let's import the right python libraries

import qiskit
from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister
from qiskit.compiler import transpile
import numpy as np
from csc_qu_tools.qiskit import Helmi

Creating the circuit

The quantum circuit is created by defining our QuantumRegister and our ClassicalRegister which hold our qubits and classical bits respectively. As this circuit only requires 2 qubits we only create a QuantumRegister of size 2. We will be measuring the outcome on each qubit and turning our quantum information into classical information, therefore we need a ClassicalRegister to hold this.

qreg = QuantumRegister(2, "qB")
creg = ClassicalRegister(2, "c")
circuit = QuantumCircuit(qreg, creg)

Now we actually add some gates to the circuit. Here a Hadamard gate is added to the first qubit or the first qubit in the quantum register. Then a Controlled-X gate is added with two arguments as it is a two qubit gate.

circuit.h(qreg[0])
circuit.cx(qreg[1], qreg[0])
circuit.measure(range(2), range(2))

Now the circuit is created! If you wish you can see what your circuit looks like by adding a print statement print(circuit.draw()) and quickly running the python script.

Decomposing the circuit

The next step is to decompose the quantum circuit you've just created into it's basis gates. These basis gates are the actual quantum gates on the quantum computer. The process of decomposition involves turning the above Hadamard and Controlled-X gates into something that can be physically run on the quantum computer. Helmi's basis gates are the two-qubit Control-Z and a the one-qubit rotational gate around the x-y plane. In Qiskit these are defined by basis_gates = ['r', 'cz'].

basis_gates = ['r', 'cz']
circuit_decomposed = transpile(circuit, basis_gates=basis_gates)

You can also print your circuit like before with print(circuit_decomposed.draw()) to see what it looks like!

Qubit Mapping

There is one more key piece of information that the Quantum Computer needs before being able to run. The Qubit Mapping. This is a python dictionary which simply states which virtual qubit should be mapped to which real qubit. The virtual qubits are the qubit's we've been using up until now, the real qubits are the ones on the quantum computer.

virtual_qubits = circuit_decomposed.qubits
qubit_mapping = {
                virtual_qubits[0]: "QB1",
                virtual_qubits[1]: "QB3",
            }

The virtual qubits are obtained from our decomposed circuit. The qubit mapping is defined via a python dictionary. Here we are mapping the first virtual qubit to the first of Helmi's qubits, QB1. The second qubit is then mapped to QB3. This is where we have made use of Helmi's topology.

The two qubit Controlled-X gate we implemented in our circuit is currently on the second of our two virtual qubits, virtual_qubits[1]. Due to Helmi's topology this needs to be mapped to QB3 on Helmi. The 1 qubit Hadamard gate can be mapped to any of the outer qubits, QB1, QB2, QB4, QB5, here we choose QB1.

Now we're ready to submit it to the Quantum Computer!

Submitting the job

First we need to set our provider and backend. The provider is the service which gives an interface to the quantum computer and the backend provides the tools necessary to submitting the quantum job.

provider = Helmi()
backend = provider.set_backend()

Before submitting the job there is one last thing we need to do: define the number of shots. The number of shots is the number of repetition of a quantum circuit. We do this because quantum computers are probabilistic machines and by repeating the result many times we can get close to a deterministic result to be able to draw conclusions from. A good number of shots for accurate results is shots = 1000.

Now we can run our quantum job!

job = backend.run(circuit_decomposed, shots=1000, qubit_mapping=qubit_mapping)

Results

Before submitting we need to ensure we can get some results! The quantum job will return what are called counts. Counts are the accumulation of results from the 1000 times the circuit is submitted to the QPU. Each time the circuit is submitted a binary state is returned, this is then added to the tally. In this case as we are submitting a 2 qubit circuit there are 4 possible resulting states: 00, 11, 01, 10. The expected results should be that approximately 50% of the counts should be in state 00 and 50% in state 11. The states of the qubits are thus entangled: if one of the qubits is measured to be in state |0>, the other one will immediately also collapse to the same state, and vice versa. As real quantum computers are not perfect, you will most likely also see that some measurements find the states |01> and |10>.

To print your results add:

counts = job.result().get_counts()
print(counts)

You can also print the entirety of job.result() which will contain all the information about your jobs results.

Save your file

Once you've made your first quantum program remember to save! CTRL+X then Y to save your file.

Running the job through LUMI

To run your quantum programme on LUMI you will need to submit the job through the SLURM batch scheduler on LUMI. Accessing Helmi is done through the q_fiqci partition. In the same directory where you have saved your quantum program, you can submit the job to SLURM using:

srun --account project_xxx -t 00:15:00 -c 1 -n 1 --partition q_fiqci python -u my-first-quantum-job.py

Remember to add your own project account!

This submits the job interactively meaning that the output will be printed straight to the terminal screen. If you wish you can also submit it using sbatch using this skeleton batch script. Using nano as before create the script batch_script.sh.

#!/bin/bash -l

#SBATCH --job-name=helmijob   # Job name
#SBATCH --output=helmijob.o%j # Name of stdout output file
#SBATCH --error=helmijob.e%j  # Name of stderr error file
#SBATCH --partition=q_fiqci   # Partition (queue) name
#SBATCH --ntasks=1              # One task (process)
#SBATCH --cpus-per-task=1     # Number of cores (threads)
#SBATCH --time=00:15:00         # Run time (hh:mm:ss)
#SBATCH --account=project_xxx  # Project for billing

module use /appl/local/quantum/modulefiles
module load helmi_qiskit

python -u my-first-quantum-job.py

This can be submitted with sbatch batch_script.sh in the same directory as your python file. Jobs in the SLURM queue can be monitored through squeue -u username and after the job has completed your results can be found in the helmijob.oxxxxx file. This can be printed to the terminal with cat.

Congratulations!

Congratulations! You have just run your first job on Helmi.

The full python script can be found below.

import qiskit
from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister
from qiskit.compiler import transpile
import numpy as np
from csc_qu_tools.qiskit import Helmi

qreg = QuantumRegister(2, "qB")
creg = ClassicalRegister(2, "c")
circuit = QuantumCircuit(qreg, creg)

circuit.h(qreg[0])
circuit.cx(qreg[0], qreg[1])
circuit.measure(range(2), range(2))

# Uncomment if you wish to print the circuit
# print(circuit.draw())

basis_gates = ['r', 'cz']
circuit_decomposed = transpile(circuit, basis_gates=basis_gates)

# Uncomment if you wish to print the circuit
# print(circuit_decomposed.draw())

virtual_qubits = circuit_decomposed.qubits
qubit_mapping = {
                virtual_qubits[0]: "QB1",
                virtual_qubits[1]: "QB3",
            }

provider = Helmi()
backend = provider.set_backend()

job = backend.run(circuit_decomposed, shots=1000, qubit_mapping=qubit_mapping)

counts = job.result().get_counts()
print(counts)

---

Last update: November 2, 2022