Nvidia releases Ising models

Quantum computers store information in qubits, which are fragile enough that heat, noise, or tiny hardware drift can spoil a calculation. Nvidia said on April 14 it is releasing Ising, an open-source family of artificial intelligence models built to tune those machines and catch their mistakes. (nvidianews.nvidia.com) Nvidia said the launch starts with two model groups: Ising Calibration, which automates quantum processor tuning, and Ising Decoding, which handles the real-time error-correction step needed while a quantum program runs. The company posted the models, training framework, and deployment guides through its developer platform on April 14. (developer.nvidia.com) Calibration is the work of repeatedly measuring a chip, adjusting control settings, and measuring again until the qubits behave as expected. Nvidia said its calibration model is a vision-language model that reads plots and lab data, then recommends the next tuning step for superconducting and spin qubit systems. (developer.nvidia.com) Error correction is the backup system quantum machines need because single qubits fail too often to be trusted on their own. Nvidia said today’s best quantum processors make an error about once every thousand operations, while useful large-scale systems would need error rates closer to one in a trillion or better. (developer.nvidia.com) Nvidia said Ising Decoding is up to 2.5 times faster and 3 times more accurate than traditional approaches, with pyMatching cited as the open-source benchmark it compared against. The company also said the models are meant to run inside hybrid systems where quantum processors work alongside graphics processing units and central processing units. (investor.nvidia.com) The release fits Nvidia’s larger push to make graphics processing units the control layer around quantum hardware rather than the quantum chip itself. Its CUDA-Q platform already lets developers split work across quantum processing units, graphics processing units, and central processing units in one program. (developer.nvidia.com) Nvidia’s quantum page says useful quantum systems will need “accelerated quantum supercomputers,” where classical machines handle orchestration, simulation, and error handling around the qubits. Ising turns part of that pitch into software that researchers can download, retrain, and plug into lab workflows now. (nvidia.com) Outside coverage described the release as developer tooling aimed at the two bottlenecks that keep quantum hardware in the lab: keeping qubits calibrated and decoding errors fast enough to correct them. Data Center Dynamics reported that the Ising family is split between calibration and decoding domains, matching Nvidia’s own rollout. (www.datacenterdynamics.com) The near-term test is whether labs beyond Nvidia can reproduce the speed and accuracy gains on their own hardware. For now, the company has put one of quantum computing’s least glamorous jobs — constant tuning and cleanup — at the center of its latest artificial intelligence release. (nvidia.com)