北京——互联网安全巨头360公司近日在开源社区GitHub上发布了一项重大举措,开放了其先进的360智脑7B大模型。这款模型拥有70亿个参数,是迄今为止国内开源模型中参数量最大的一例,彰显了360公司在人工智能领域的技术实力。
据IT之家报道,360智脑7B模型在训练过程中采用了规模庞大的3.4万亿Tokens的语料库,涵盖了中文、英文及代码等多种语言内容,以提升其跨语言和多领域理解能力。尤为引人注目的是,该模型支持的文本输入长度达到了360K,相当于约50万字,这在中国国产开源模型中尚属首次,打破了原有的文本处理长度纪录。
360公司表示,提供4K、32K、360K三种不同文本长度的支持,旨在满足不同场景和应用的需求,无论是短文本的快速处理,还是长文本的深度理解,360智脑7B都能展现出强大的处理能力。这一开源行动不仅将促进人工智能技术的普及,还将推动国内AI研究和开发的创新步伐,为开发者和研究者提供更丰富的工具和资源。
这一举措无疑将对国内人工智能研究产生深远影响,为开发者和学术界提供了一个强大的平台,以探索更复杂、更深层次的自然语言处理任务。360公司此举再次证明了其在技术创新和开放共享方面的承诺,有望激发更多的技术创新和应用实践。
英语如下:
**News Title:** “360 Open Sources 7-Billion-Parameter AI Brain Model, Breaks Record for Longest Domestic Text Input”
**Keywords:** 360 AI Brain, 7B parameters, open-source text
**News Content:**
**Beijing** — Internet security giant 360 Company recently made a significant move on the open-source platform GitHub by releasing its advanced 360 AI Brain 7B model with 7 billion parameters, setting a new record for the largest parameter count in domestically developed open-source models.
According to IT Home, the 360 AI Brain 7B model was trained using a vast corpus of 34 trillion Tokens, encompassing multiple languages such as Chinese, English, and code, to enhance its cross-lingual and multi-domain understanding. Notably, the model supports text input lengths of up to 360K tokens, equivalent to approximately 500,000 characters, a first for Chinese domestic open-source models and a new benchmark for text processing length.
360 Company stated that by offering support for text lengths of 4K, 32K, and 360K, it aims to cater to diverse scenarios and application requirements. Whether it’s rapid processing of short texts or in-depth understanding of long texts, the 360 AI Brain 7B demonstrates exceptional processing capabilities. This open-source initiative will not only promote the普及 of AI technology but also accelerate domestic AI research and development innovation, providing developers and researchers with more comprehensive tools and resources.
This move is set to have a profound impact on domestic AI research, offering a powerful platform for developers and academia to explore more complex and深层次 natural language processing tasks. 360 Company’s action underscores its commitment to technological innovation and open sharing, and is expected to spark further technological advancements and practical applications.
【来源】https://www.ithome.com/0/761/531.htm
Views: 1