今日下午，OPPO 举行了 Find X7 系列产品技术沟通会，公布了安第斯大模型 AndesGPT 的全新升级。作为一款拥有 70 亿参数的大模型，AndesGPT 在 Find X7 系列手机上首次实现了端侧应用。此次升级，OPPO 针对不同场景的用户需求，将安第斯大模型分为 1800 亿参数、700 亿参数、70 亿参数三种体量，实现智能调度。
为了在 Find X7 系列上实现 70 亿参数大模型的端侧应用，OPPO 采取了多种举措。首先，通过 4bit 量化的模型压缩，减小了模型体积，提高了运算效率。其次，OPPO 优化了 Al Boost 推理引擎，提升了推理速度。最后，OPPO 与芯片厂家深度合作，针对大模型进行了联合优化。
此次 AndesGPT 全新升级，标志着 OPPO 在人工智能领域取得了重要突破。未来，OPPO 将继续探索端侧大模型的应用，为用户提供更智能、更便捷的手机体验。
News content:
At today’s OPPO Find X7 series product technology communication meeting, OPPO announced the new upgrade of AndesGPT, a large-scale model with 7 billion parameters. As a large model with 7 billion parameters, AndesGPT has achieved the first-time application on the terminal side of the OPPO Find X7 series. To meet the different user needs in various scenarios, OPPO has divided AndesGPT into three sizes: 180 billion parameters, 700 billion parameters, and 70 billion parameters, realizing intelligent dispatching.
To achieve the terminal application of the 70-billion-parameter large model on the OPPO Find X7 series, OPPO has taken various measures. Firstly, through 4bit quantization of the model compression, the model volume is reduced and the operation efficiency is improved. Secondly, OPPO optimizes the Al Boost inference engine, enhancing the inference speed. Finally, OPPO deeply cooperates with chip manufacturers to jointly optimize the large model.
The new upgrade of AndesGPT marks an important breakthrough for OPPO in the field of artificial intelligence. In the future, OPPO will continue to explore the application of large-scale models on the terminal side, providing users with more intelligent and convenient mobile phone experiences.
【来源】https://www.ithome.com/0/741/887.htm
Views: 1