As quantum computers scale, they will integrate with AI supercomputers to tackle some of the world’s most challenging problems. These accelerated quantum supercomputers will run applications leveraging the capabilities of CPUs, GPUs, and QPUs.
With the NVIDIA CUDA-Q platform, users can easily perform research and develop applications that can seamlessly run on accelerated quantum supercomputers, integrating any qubit modality, as well as in simulation. CUDA-Q is currently being used to develop diverse applications across academia and industry, including simulating better quantum hardware, studying error correction, and more.
There is an increasing need for educational resources to teach students how to work with tools capable of supporting such hybrid environments within accelerated quantum supercomputers. NVIDIA CUDA-Q Academic is designed to bridge this gap combining theory and practice to prepare the next generation of quantum computing (QC) professionals to work with accelerated quantum supercomputers using CUDA-Q.
CUDA-Q Academic is a freely available collection of interactive Jupyter notebooks developed in collaboration with numerous partner universities and tested in real classroom settings. The modular, interactive lessons include video explanations, exercises, and solutions that give students hands-on experience programming with CUDA-Q.
“Collaborative efforts between industry and academia are essential for advancing quantum computing and high-performance computing education,” said Ramin Ayanzadeh, an assistant professor of Computer Science at University of Colorado Boulder who is helping to shape the curriculum. “These partnerships ensure that students are exposed to the latest technological developments and practical applications, preparing them to address real-world challenges in hybrid quantum-classical systems.”
Two distinct collaborations between the Neils Bohr Institute at the University of Copenhagen and NVIDIA highlight the value and diversity of such partnerships. The institute will use CUDA-Q as the primary platform to leverage AI supercomputers like Gefion and build course material around large-scale simulation of quantum systems. Simultaneously, Professor Gemma Solomon is spearheading work to build focused CUDA-Q Academic content for introducing chemistry students to the principles of quantum computing.
This post provides an overview of CUDA-Q Academic and how it can be used to build quantum programming skills that are useful today and well into the era of large-scale accelerated quantum supercomputers.
Practical quantum computing preparation with CUDA-Q Academic
CUDA-Q Academic focuses on the realities of practical quantum computing by combining high-performance computing skills (HPC) with QC skills. Students are introduced to hybrid quantum-classical workflows and GPU acceleration for large-scale problems. This approach is already making an impact in the classroom, as expressed by instructor Daniel Justice at Carnegie Mellon University: “After teaching Quantum Computing for five years at CMU, introducing CUDA-Q was a game-changer. For the first time, my students had access to an interactive quantum application powered by GPU-accelerated simulation, enabling them to work with a number of qubits and tackle problem sizes previously unmanageable in our curriculum.”
CUDA-Q Academic offers multiple tracks for students to create a customized learning plan based on their background and interests (Figure 1). The modules begin with textbook examples and progress to more advanced techniques directly from current research literature.
Beginners are recommended to start with the Quick Start to Quantum Computing with CUDA-Q track, which takes learners from the definition of a qubit and a quantum Hello World program to implementing a variational algorithm on a GPU in less than a day. This track lays a firm foundation for more advanced topics, such as the Quantum Applications to Finance module, which features cutting-edge research on multi-split-step quantum walks by Ching-Ray Chang, professor and director of the CYCU Quantum Information Center, and his team.
“The Quick Start to Quantum Computing with CUDA-Q track has not only provided students in my Quantum Computation class with a strong foundation for quantum computing, but quickly prepares them to interact with the latest quantum algorithm research,” said Junyu Liu, assistant professor of Computer Science at University of Pittsburgh.
After completing the Quick Start track, students are prepared to tackle any of the other CUDA-Q Academic tracks, which range from quantum error correction to solving large-scale optimization problems with circuit cutting techniques.
Example CUDA-Q Academic module: QAOA for Max Cut
Implementing a divide-and-conquer approach with the Quantum Approximate Optimization Algorithm (QAOA) to solve the Max Cut problem offers an engaging and visually intuitive way to understand circuit cutting techniques. This is one of many strategies to accelerate quantum circuit simulation on GPUs and to distribute quantum workloads across multiple QPUs.
The QAOA for Max Cut series demonstrates the layout of CUDA-Q Academic modules. Students begin with Notebook 0, which provides detailed instructions for preparing the learning environment, ensuring a seamless experience on platforms like CoCalc or qBraid.
Next, students proceed to Notebook 1, which features interactive tools for exploring and experimenting with the concepts, enhancing their understanding and engagement with the material.
Additional notebooks then build up the theory and programming fundamentals for students to prepare their own solutions.
Teaching HPC and quantum computing concepts together
A feature that distinguishes CUDA-Q Academic modules from other quantum educational content is their transparent approach to the challenges involved in quantum algorithm implementations and how HPC tools can help overcome them.
One common challenge in practical quantum computing, for example, is the limited number of qubits. Most problems of interest require far more qubits than currently are available, whether through execution on existing quantum computers or through simulation on GPUs.
In the example of the Max Cut module, later notebooks teach students how to address this practical challenge with circuit cutting. Circuit cutting is an approach that breaks down quantum circuits into smaller circuits, each of which may require fewer qubits than the original circuit (Figure 2). Often the smaller circuits can be executed in parallel before their output is merged back together with an approximation of the original circuit execution.
Exploring work like circuit cutting requires access to the hybrid resources offered by CUDA-Q. In Notebook 2 of the Max Cut module, a divide-and-conquer approach is introduced wherein students are shown how to leverage GPUs to simulate multiple QPUs working in tandem to implement circuit cutting.
CUDA-Q Academic modules also allow students to grapple with the practicalities of implementing quantum algorithms. For example, in Notebook 3 of the Max Cut module, students have an opportunity to experiment with many of the design decisions that researchers face when implementing QAOA and circuit cutting. In addition to learning the QAOA algorithm, learners gain transferable skills in HPC as they simulate large-scale algorithms on a GPU using Message Processing Interface (MPI).
Students learn actively through the entire module with numerous coding exercises accompanied by detailed solutions. After completing the three lab notebooks, students are challenged with a final assessment lab. In this final lab, students apply all they have learned to an implementation of the weighted Max Cut problem.
Completing the Max Cut lab not only introduces new quantum concepts, but also demonstrates how HPC can be integrated with quantum computing. It prepares students to understand advanced applications, such as QAOA-GPT and Adaptive Circuit Knitting, by showcasing the role of GPUs in today’s cutting-edge technologies.
Getting started with quantum computing is easier than ever
For students and professionals interested in developing HPC and QC skills, the Jupyter notebooks—along with instructions about how to get started—are available at the CUDA-Q Academic GitHub repository. Seamless integrations with platforms like CoCalc and qBraid are also available.
For instructors interested in adding CUDA-Q materials to your curriculum, check out the CUDA-Q Sample Syllabus. You can also leverage the learning management and collaboration tools in CoCalc to share the CUDA-Q Academic materials with your students.
Interested in co-developing or pilot-testing CUDA-Q Academic lessons? Visit NVIDIA Quantum to sign up for more information.
