lightning.amdgpu

The lightning.amdgpu is a specific instance of our lightning.kokkos device to give maximum performance on AMD GPUs.

Recommended for:

• Large state vector simulations with 20+ qubits on AMD GPUs (e.g., on AMD Developer Cloud).
• Fast gradients using the adjoint differentiation method.
• Linux operating systems.
• Distributed state vector support with Message Passing Interface (MPI) is also supported (through lightning.kokkos).

Documentation

To learn more, please visit the device documentation:

• lightning.amdgpu documentation

See all Lightning devices:

• lightning.qubit
• lightning.kokkos
• lightning.amdgpu
• lightning.gpu
• lightning.tensor

Installation

For details on installation and dependencies, visit the lightning.amdgpu installation page. For instructions to build for other architectures or with MPI support, visit the lightning.kokkos installation page

Device Initialization

Initialize the device in PennyLane with:

import pennylane as qml
dev = qml.device("lightning.amdgpu", wires=20)

For more details on device settings and keyword arguments, see the device documentation for lightning.amdgpu and lightning.kokkos.