PennyLane-Qrack

Usage

API

  1. Docs
  2. The Simulator device

The Simulator device

You can instantiate the Qrack device in PennyLane as follows:

import pennylane as qml
from catalyst import qjit # optional, for JIT compilation

dev = qml.device('qrack.simulator', wires=2)

This device can then be used just like other devices for the definition and evaluation of QNodes within PennyLane. A simple quantum function that returns the expectation value of a measurement and depends on three classical input parameters would look like:

@qjit # optional, for JIT compilation
@qml.qnode(dev)
def circuit(x, y, z):
    qml.RZ(z, wires=[0])
    qml.RY(y, wires=[0])
    qml.RX(x, wires=[0])
    qml.CNOT(wires=[0, 1])
    return qml.expval(qml.PauliZ(wires=1))

You can then execute the circuit like any other function to get the quantum mechanical expectation value.

circuit(0.2, 0.1, 0.3)

Device options

By default, Qrack will automatically and “transparently” switch between CPU-based and GPU-based methods as would be most appropriate for best performance. Qrack will also automatically use the “hybrid simulation algorithm” “layers” that tend to work best for general “BQP-complete” tasks like random circuit sampling.

If you’d like to use a special-case algorithm, like quantum binary decision diagrams (QBDD), or if you want more direct control over choice of algorithm and hardware used, there is a series of boolean options users can toggle like so:

dev = qml.device('qrack.simulator', wires=2, isBinaryDecisionTree=True)

This is a complete list of boolean options and their meanings:

Parameter

Default

Description

is_stabilizer_hybrid

True

Use “hybrid” stabilizer optimization? (Non-Clifford circuits will fall back to near-Clifford or universal simulation.)

is_schmidt_decompose_multi

False

Distribute Schmidt-decomposed qubit subsystems to multiple GPUs or accelerators? (Mismatched device capacities might hurt overall performance.)

is_binary_decision_tree

False

Use “quantum binary decision diagram” (“QBDD”) methods? (Note that QBDD is CPU-only.)

is_gpu

True

Use GPU acceleration?

is_near_clifford_tableau_writer

False

Use multi-GPU (or multi-memory-segment) acceleration?

is_host_pointer

False

Allocate GPU buffer from general host heap? (Might improve performance or reliability, like when accelerating on an Intel HD)

is_sparse

False

For CPU-based simulation, use sparse state vectors (Default is “false”)

noise

0

Depolarizing noise parameter (Noise intensity also responds to “QRACK_GATE_DEPOLARIZATION” environment variable)

Supported operations

The qrack.simulator device supports all PennyLane operations and observables.

Previous
Next