Posted On: May 5, 2022
Amazon Braket, the quantum computing service from AWS, is designed to make it easy for customers to conduct scientific research and software development with quantum computers. Today, we are excited to announce that Amazon Braket Hybrid Jobs, the capability to combine quantum and classical compute to explore hybrid algorithms, now supports embedded circuit simulators, improving performance by up to 10X or more. By bringing the quantum simulator closer to your classical algorithm you can reduce the simulator round-trip time compared to using Braket’s on-demand circuit simulators. This new capability can help reduce a performance bottleneck when running large volumes of quantum simulation tasks that each use only a relatively small number of qubits, typically up to ~25 qubits.
Prior to this launch, you could choose to use one of Amazon Braket’s five quantum processing units (QPU), or three fully-managed, on-demand quantum circuit simulators: SV1, DM1, or TN1. Now, for hybrid jobs, you can also choose from five high-performance, embedded simulators from PennyLane, such as the fast lightning.cpu state-vector simulator or the lightning.gpu simulator accelerated using the NVIDIA cuQuantum SDK, that come pre-installed in jobs containers. In addition to reducing latency, these embedded simulators support advanced capabilities, such as the adjoint differentiation method for more efficient gradient computations, enabling even faster algorithm iteration and innovation with hybrid algorithms. For data-intensive jobs such as quantum machine learning algorithms, embedded simulators also support the ability for you to execute circuits in parallel across multiple CPU or GPU instances by only changing a few lines of code. By distributing your job across multiple instances, you can take advantage of the elasticity of the cloud to reduce algorithm runtimes even further, while still only paying for what you use.
The use of embedded circuit simulators is available in all of the regions where Amazon Braket is available. To get started with embedded simulators, refer to this example notebook. To learn more about how Amazon Braket Hybrid Jobs makes it easier and faster to run quantum machine learning workloads, refer to this blog.