【360公司开源70亿参数大模型，刷新国产开源模型纪录】近日，互联网安全巨头360公司宣布在知名代码托管平台GitHub上开源其最新的360智脑7B大模型。这款模型拥有70亿参数，是迄今为止国内开源模型中参数量最大的一款，展示了360公司在人工智能领域的技术实力。
据360公司透露，360智脑7B大模型基于一个规模庞大的3.4万亿Tokens的语料库进行训练，涵盖了中文、英文以及代码等多种语言内容，旨在实现更为精准和全面的语言理解和生成能力。值得关注的是，该模型支持4K、32K和360K三种不同长度的文本输入，其中360K的文本长度约相当于50万字，这一长度打破了国产开源模型的记录，为开发者提供了更广阔的应用场景。
这一开源举措将极大地促进国内AI研究和开发的进程，为学术界和工业界提供了一个强大的工具，有助于推动自然语言处理技术的进步。360公司此举不仅彰显了其在AI领域的开放精神，也预示着国产AI技术将有更多创新和突破的可能性。对于开发者来说，360智脑7B大模型的开源将为他们提供一个宝贵的实验平台，以实现更高效、更智能的文本处理应用。
英语如下：
News Title: “360 Open Sources 7 Billion-Parameter Intelligent Brain Model, Breaking Domestic Record for Longest Text Input!”
Keywords: 360 Intelligent Brain, 7B Parameters Open Source, Longest Text
News Content: **360 Company Open Sources 7 Billion-Parameter Large Model, Setting a New Record for Domestic Open-Source Models**
Recently, internet security giant 360 Company announced the open sourcing of its latest 360 Intelligent Brain 7B large model on the renowned code hosting platform GitHub. With 7 billion parameters, this model is the largest in terms of parameter count among domestic open-source models, showcasing 360’s technological prowess in the field of artificial intelligence.
According to 360 Company, the 360 Intelligent Brain 7B model has been trained on an enormous corpus of 3.4 trillion Tokens, encompassing Chinese, English, and code in multiple languages. Its purpose is to achieve more precise and comprehensive language understanding and generation capabilities. Notably, the model supports text inputs of 4K, 32K, and 360K tokens in length, with the 360K length equivalent to approximately 500,000 characters, breaking the record for the longest input length in domestic open-source models and offering developers a broader range of application scenarios.
This open-source initiative will significantly advance the domestic AI research and development process, providing the academic and industrial sectors with a powerful tool to contribute to the progress of natural language processing technology. By doing this, 360 Company not only demonstrates its open spirit in the AI domain but also signals the potential for more innovation and breakthroughs in domestically developed AI technologies. For developers, the open sourcing of the 360 Intelligent Brain 7B model presents a valuable experimental platform to develop more efficient and intelligent text processing applications.
【来源】https://www.ithome.com/0/761/531.htm
Views: 1