05/04/2026
Origin Quantum's self-developed quantum machine learning framework, VQNet 2.17, focuses on “quantum-classical integration" for computing power synergy, providing an efficient and easy-to-use industrialization solution.
The quantum machine learning framework VQNet, independently developed by Origin Quantum, has been upgraded to version 2.17, centering on “quantum-classical integration" to deliver an efficient and user-friendly solution for industrial deployment. The new version significantly improves hybrid training efficiency in Linux GPU environments, saving users time costs. By restructuring distributed communication operators and optimizing video memory management, it enables training of 30qubit variational quantum circuits using just two consumergrade 24 GB graphics cards, breaking through hardware limitations faced by similar products. The team also innovatively proposed a quantumtensor hybrid parameter finetuning method, integrating the highdimensional representation capability of quantum parameter spaces into a hundredbillionparameter classical large model. This method outperforms classical models of the same scale on professional tasks, validating the application potential of “quantum + AI.” The efficient capabilities of VQNet 2.17 can broadly support fields such as autonomous driving, natural language processing, financial modeling, and drug discovery, accelerating the transition of quantum technology from laboratories to industries.