Hi everyone! We’re very happy to announce the release of PennyLane version 0.10
New and improved simulators
-
Use the new default.qubit.tf device, a quantum simulator written in TensorFlow, allowing simulated circuits to be trained faster via backpropagation
-
Experiment with an upgraded default.tensor plugin that now features two different tensor network representations to be used:
"exact"
and"mps"
New machine learning functionality and integrations
-
Convert a PennyLane QNode into a TorchLayer, allowing for creation of quantum and hybrid models using the
torch.nn
-
Trade off extra computation for enhanced memory efficiency with the new “reversible” differentiation method which can be used in simulators
New templates using broadcasting
-
Use the new
UCCSD
,SingleExcitationUnitary
, andDoubleExcitationUnitary
templates to create the Unitary Coupled-Cluster Singles and Doubles (UCCSD) ansatz and perform VQE-based quantum chemistry simulations using PennyLane-QChem -
Specify the Squared error loss function for circuits with trainable parameters using the
SquaredErrorLoss
class from the newpennylane.qnn.cost
module
Major upgrades to variable handling
Mark if a variable is differentiable or not by using the new pennylane.numpy.tensor
class and specifying the requires_grad
attribute
New plugin to complement the growing library of devices
In addition to @qiskit, @rigetti Forest, @GoogleAI Cirq, etc., you can now access Alpine Quantum Technologies’ online ion-trap platform using PennyLane.
The full release notes are available at Release notes — PennyLane 0.27.0 documentation
As always, this release would not have been possible without all the help from our contributors:
@Alain_Delgado_Gran, @antalszava , Jack Ceroni , @Maria_Schuld, Nicola Vitucci, @theodor, @josh, @nathan , @Tom_Bromley .