Quantum-computing firm D-Wave has introduced an open-source toolkit for bringing quantum annealing into AI and machine learning (ML) workflows.

The toolkit, part of D-Wave’s Ocean software suite, integrates directly with PyTorch, one of the most widely used frameworks for building and training ML models, bringing quantum processing to a framework typically tied to classical computing.

Developers will be able to train restricted Boltzmann machines (RBMs), a type of neural network often used in generative AI (genAI), using D-Wave’s annealing-based quantum processors. Quantum annealing is capable of optimizing feature selection, which D-Wave argues is the fundamental ML building block.

With the PyTorch integration relying on the cloud-based Ocean platform, developers can trial quantum workloads over the network, much like renting GPUs for deep learning.

What is quantum annealing?

Quantum annealing uses quantum fluctuations to find the best solution, while gate-model quantum computers offered by IBM and other quantum computing vendors require problems to be expressed in terms of quantum gates.

With its toolkit, D-Wave is demonstrating that annealing is ready for practical AI tasks, meaning developers don’t need to wait before gate-based machines arrive at scale.

The move is in line with other offerings from D-Wave that bring quantum experimentation in the cloud to developers. Its flagship Leap cloud platform offers Quantum-Computing-as-a-Service, with integration available on Amazon Web Services (AWS) Marketplace.

Strengthening its annealing position, the firm announced the first commercial sale of its Advantage quantum computer earlier this year to an undisclosed customer, broadening its overall go-to-market offering to include on-premises system sales.

D-Wave also claimed recently that it had demonstrated quantum supremacy using its Advantage2 prototype annealing quantum computer – although experts were skeptical of the announcement.