【北京大学开源全新aiXcoder-7B代码大模型，引领企业级软件开发新高度】
近日，北京大学的aiXcoder团队宣布开源其最新研发的aiXcoder-7B Base版代码大模型，这是一款专为企业私有部署设计的先进工具，旨在提升企业软件开发的效率与质量。该模型在海量的1.2T Unique Tokens数据集上进行了深度训练，其设计和预训练任务均针对真实的代码生成场景，展现了极高的专业性和针对性。
据量子位报道，aiXcoder-7B Base在HumanEval、MBPP和MultiPL-E三大主流代码评测集上的表现超过了拥有340亿参数的Codellama模型，彰显了其在代码补全场景中的优越性能。在多语言NL2Code基准测试中，aiXcoder-7B Base的平均表现同样超越了Codellama 34B和StarCoder2 15B，显示出其在处理复杂编程任务时的领先能力。
这一创新性的开源项目不仅体现了北京大学在人工智能领域的科研实力，也为全球企业提供了更高效、更智能的代码生成解决方案。aiXcoder-7B Base的推出，无疑将推动企业软件开发领域的技术革新，助力企业提升开发效率，降低开发成本，同时为未来的代码智能生成领域设立了新的标准。
英语如下：
News Title: “Peking University Launches Open-Source aiXcoder-7B: A New Benchmark in Enterprise Code Generation!”
Keywords: aiXcoder-7B, Enterprise Deployment, Code Generation
News Content: **Peking University Unveils Open-Source aiXcoder-7B Large Language Model, Elevating Enterprise Software Development**
Recently, the aiXcoder team at Peking University announced the open-source release of their cutting-edge aiXcoder-7B Base code generation model, designed specifically for enterprise-private deployment. This advanced tool aims to enhance efficiency and quality in corporate software development. Trained extensively on a massive 1.2T Unique Tokens dataset, the model’s design and pre-training tasks are tailored for real-world code generation scenarios, demonstrating exceptional professionalism and focus.
As reported by Quantum Bit, aiXcoder-7B Base outperforms Codellama, a model with 340 billion parameters, across the major code evaluation benchmarks HumanEval, MBPP, and MultiPL-E, showcasing its superior performance in code completion tasks. In the multilingual NL2Code benchmark, aiXcoder-7B Base also surpasses Codellama 34B and StarCoder2 15B, indicating its leading capabilities in handling complex programming assignments.
This innovative open-source project not only highlights Peking University’s research prowess in artificial intelligence but also offers global enterprises a more efficient and intelligent code generation solution. The introduction of aiXcoder-7B Base is set to drive technological innovation in enterprise software development, boosting efficiency, and reducing costs. It also establishes a new standard for the future of intelligent code generation.
【来源】https://www.qbitai.com/2024/04/134070.html
Views: 1