上海岩芯数智人工智能科技有限公司（岩芯数智，RockAI）于1月24日下午宣布，成功研发并发布国内首款自研的非Transformer Attention机制的通用自然语言大模型——Yan模型。该模型在记忆能力、处理速度以及推理吞吐量等方面均有显著提升，分别达到原有水平的3倍、7倍和5倍。
据悉，Yan模型是我国首个发布的与ChatGPT不同机制的通用大模型，其参数规模达到百亿。该公司表示，这一模型实现了用百亿级参数达成千亿参数大模型的性能效果。
Yan模型的发布标志着我国在自然语言处理领域取得了重要突破，将进一步推动人工智能技术的发展和应用。岩芯数智作为国内AIGC领域的领军企业，其成功研发的Yan模型有望为各行各业带来更智能、更高效的解决方案。
英文标题：RockAI Unveils China’s First Self-developed Non-Transformer Attention Large Model Yan
关键词：RockAI, Non-Transformer Attention, Yan Model
新闻内容：
Shanghai RockAI, a leading AI company in China, has announced the successful development and release of China’s first self-designed general-purpose natural language large model with a non-Transformer Attention mechanism, named Yan model. The new model has significantly improved in terms of memory capacity, processing speed, and inference throughput, achieving 3 times, 7 times, and 5 times enhancements respectively over the previous versions.
It is understood that Yan model is the first general-purpose large model in China with a different mechanism from ChatGPT, boasting a parameter scale of 100 billion. The company claims that this model achieves the performance of a 1000 billion parameter large model with 100 billion parameters.
The release of Yan model signifies an important breakthrough in the field of natural language processing in China and is expected to further promote the development and application of artificial intelligence technology. As a leading company in the AIGC sector, RockAI’s successful development of the Yan model is expected to bring more intelligent and efficient solutions to various industries.
【来源】https://www.tmtpost.com/6898099.html
Views: 1