This Python package uses a backend written in Julia to implement high performance features for standard Qiskit. This package is a proof of concept with little high-level code.
Installing and managing Julia and its packages is automated. So you don't need to learn anything about Julia to get started.
Installation and configuration notes
Compilation Compiling a system image to eliminate compilation at run time.
Using qiskit_alt First steps.
Manual Steps Details of automatic installation.
Julia Packages Julia packages that qiskit_alt depends on.
Development. Instructions for developing qiskit_alt.
qiskit_alt
is available on pypishell> pip install qiskit_alt
import qiskit_alt
qiskit_alt.project.ensure_init()
See julia_project
for more options.
If no Julia executable is found, jill.py
will be used to download and install it. It is not necessary
to add the installation path or symlink path to your search PATH to use julia from qiskit_alt.
Before offering to install Julia, qiskit_alt
will search for julia as described here.
The Julia packages are installed the first time you run qiskit_alt.project.ensure_init()
from Python. If this fails,
see the log file qiskit_alt.log. You can open a bug report in the qiskit_alt
repo
Check that the installation is not completely broken by running benchmark scripts, with the string "alt" in the name:
python ./bench/run_all_bench.py
Note that the folder bench
is not included when you install via pip install qiskit_alt
.
But it can be downloaded here.
qiskit_alt
depends on pyjulia
and/or juliacall
for communication between Julia and Python.
pyjulia
and juliacall
are two packages for communication between Python and Julia. You only need
to import one of them. But, you won't import them directly.
When you initialize with qiskit_alt.project.ensure_init()
the default communication package is chosen.
You can also choose explicitly with qiskit_alt.project.ensure_init(calljulia="pyjulia")
or qiskit_alt.project.ensure_init(calljulia="juliacall")
The installation is interactive. How to do a non-interactive installation with environment variables is described below.
You may allow qiskit_alt
to download and install Julia for you, using jill.py
.
Otherwise you can follow instructions for installing Julia with an installation tool.
We recommend using a virtual Python environment with venv
or conda
. For example python -m venv ./env
,
which creates a virtual environment for python packages needed to run qiskit_alt
.
You can use whatever name you like in place of the directory ./env
.
Activate the environment using the file required for your shell. For example
source ./env/bin/activate
for venv
and bash.
Install qiskit_alt
with pip install qiskit_alt
.
Install whatever other packages you want. For example pip install ipython
.
Configuring installation with environment variables. If you set these environment variables, you will not be prompted during installation.
QISKIT_ALT_JULIA_PATH
to the path to a Julia executable (in a Julia installation). This variable takes
precedence over other methods of specifying the path to the executable.QISKIT_ALT_INSTALL_JULIA
to y
or n
to confirm or disallow installing julia via jill.py
.QISKIT_ALT_COMPILE
to y
or n
to confirm or disallow compiling a system image after installing Julia packagesQISKIT_ALT_DEPOT
to y
or n
to force using or not using a Julia "depot" specific to this project.qiskit_alt.project.update()
will delete Manifest.toml
files; upgrade packages; rebuild the manifest; delete compiled system images.
If you call update()
while running a compiled system image, you should exit Python and start again before compiling
qiskit_alt.project
is an instance of JuliaProject
from the package julia_project
.
for managing Julia dependencies in Python projects. See more options at julia_project
.
qiskit_alt
to speed up loading and reduce delays due to just-in-time compilation.
You will be prompted to install when installing or upgrading. Compilation may also be done at any time as follows.[1]: import qiskit_alt
In [2]: qiskit_alt.project.ensure_init(use_sys_image=False)
In [3]: qiskit_alt.project.compile()
Compilation takes about four minutes. The new Julia system image will be found the next time you import qiskit_alt
.
Note that we disabled possibly loading a previously-compiled system image before compiling a new one.
This avoids some possible stability issues.
This is a very brief introduction.
The pyjulia
interface is exposed via the julia
module. The juliacall
module is called juliacall
.
However you should not do import julia
or import juliacall
before import qiskit_alt
,
and qiskit_alt.project.ensure_init()
(or qiskit_alt.project.ensure_init(calljulia="pyjulia")
or
juliacall
with qiskit_alt.project.ensure_init(calljulia="juliacall")
)
This is because import julia
will circumvent the facilities described above for choosing the julia executable and the
compiled system image.
Julia modules are loaded like this. Note that qiskit_alt.project.julia
points to either julia
or juliacall
depending
on which was chosen.
import qiskit_alt
qiskit_alt.project.ensure_init(calljulia=interface_choice)
Main = qiskit_alt.project.julia.Main
import qiskit_alt
; import julia
; from julia import PkgName
.
After this, all functions and symbols in PkgName
are available.
You can do, for example
In [1]: import qiskit_alt
In [2]: qiskit_alt.project.ensure_init()
In [3]: julia = qiskit_alt.project.julia
In [4]: julia.Main.cosd(90)
Out[4]: 0.0
In [5]: QuantumOps = qiskit_alt.project.simple_import("QuantumOps")
In [6]: pauli_sum = QuantumOps.rand_op_sum(QuantumOps.Pauli, 3, 4); pauli_sum
Out[6]:
<PyCall.jlwrap 4x3 QuantumOps.PauliSum{Vector{Vector{QuantumOps.Paulis.Pauli}}, Vector{Complex{Int64}}}:
IIZ * (1 + 0im)
XYI * (1 + 0im)
YIX * (1 + 0im)
ZIZ * (1 + 0im)>
In the last example above, PauliSum
is a Julia object. The PauliSum
can be converted to
a Qiskit SparsePauliOp
like this.
In [7]: from qiskit_alt.pauli_operators import PauliSum_to_SparsePauliOp
In [8]: PauliSum_to_SparsePauliOp(pauli_sum)
Out[8]:
SparsePauliOp(['ZII', 'IYX', 'XIY', 'ZIZ'],
coeffs=[1.+0.j, 1.+0.j, 1.+0.j, 1.+0.j])
The highlights thus far are in benchmark code, which is
presented in the demonstration notebooks.
There is one demonstration notebook using pyjulia
and another demonstration notebook using juliacall
.
The main application-level demonstration is computing a qubit Hamiltonian as a qiskit.quantum_info.SparsePauliOp
from a Python list specifiying the molecule geometry in the same format as that used by qiskit_nature
.
qiskit.quantum_info.SparsePauliOp
,
is many times faster with the factor increasing rapidly in the number of qubits.You might want to skip to installation instructions
There are a few demos in this demonstration benchmark notebook
The benchmark code is a good place to get an idea of what qiskit_alt can do.
Here are some demonstration notebooks of the Julia packages behind qiskit_alt
.
Zapata demo of Jordan-Wigner transformation in Julia; The same thing as the main demonstration in qiskit_alt. This is from JuliaCon 2020.
Project.toml
) by doing qiskit_alt.project.julia.Pkg.add("PackageName")
.
You can also do the same by avoiding Python and using the julia cli.QuantumOps.jl
and
ElectronicStructure.jl
and
QiskitQuantumInfo.jl
are not registered in the General Registry, but rather in QuantumRegistry
which contains just
a handful of packages for this project.The test folder is mostly out of date.
See the readme.