近日，国内AIGC企业上海岩芯数智人工智能科技有限公司(岩芯数智，RockAI)发布了国内首个自研的非Transformer Attention机制的低算力通用自然语言大模型——Yan模型。据悉，该模型的记忆能力提升了3倍，速度提升了7倍，推理吞吐量提升了5倍。这是国内首个发布的与ChatGPT不同机制的通用大模型，参数规模达到了百亿级。
Yan模型的研发过程中，岩芯数智采用了非Transformer Attention机制，这种机制相较于传统的Transformer Attention机制具有更高的效率和更低的计算成本。同时，Yan模型还具备强大的自然语言理解能力和生成能力，可以广泛应用于智能客服、智能问答、机器翻译等领域。
岩芯数智表示，他们通过不断地优化和改进算法，最终实现了用百亿级参数达成千亿参数大模型的性能效果。这一突破性的成果将为中国人工智能产业的发展注入新的活力和动力。
目前，Yan模型已经正式上线并开始接受用户测试。未来，岩芯数智将继续深耕人工智能领域，推出更多具有创新性和实用性的产品和服务，为推动中国人工智能产业的发展做出更大的贡献。
英语如下：
Title: “Yan” Model Released: China’s First Self-Developed Large Model with Non-Transformer Attention Mechanism, Memory Ability Increases by 3 Times, Speed Increases by 7 Times!
Keywords: Non-Transformer Attention, Low-computational-power General Model, Yan Core Artificial Intelligence Release
Recently, Shanghai RockAI Artificial Intelligence Technology Co., Ltd. (RockAI), an AIGC enterprise in China, released the country’s first self-developed large natural language model with a non-Transformer Attention mechanism – Yan model. It is reported that this model has a memory ability of three times increased and a speed increase of seven times, with an inference throughput increase of five times. This is the first general large model released in China that differs from the ChatGPT mechanism, with a parameter scale reaching hundreds of billions.
In the process of developing Yan model, RockAI adopted the non-Transformer Attention mechanism, which相比于 traditional Transformer Attention mechanism has higher efficiency and lower computational cost. At the same time, Yan model also has strong natural language understanding and generation capabilities and can be widely applied in intelligent customer service, intelligent question answering, machine translation, and other fields.
RockAI said that they continuously optimized and improved the algorithm to achieve the performance effect of a thousand-billion-parameter large model with one hundred billion parameters. This groundbreaking achievement will inject new vitality and momentum into the development of China’s artificial intelligence industry.
At present, Yan model has been officially launched and is beginning to receive user testing. In the future, RockAI will continue to delve into the field of artificial intelligence, launch more innovative and practical products and services, and make greater contributions to promoting the development of China’s artificial intelligence industry.
【来源】https://www.tmtpost.com/6898099.html
Views: 1