【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团队在开源领域再创佳绩，正式发布了专为企业私有部署设计的aiXcoder-7B Base版代码大模型。这款模型以其强大的性能和精准的代码生成能力，为软件开发领域带来了新的突破。
据量子位报道，aiXcoder-7B Base模型在1.2T Unique Tokens的海量数据上进行了深度训练，其设计目标直指真实的代码生成场景。在预训练任务和上下文信息处理上，该模型展现出了独特的匠心，旨在提升企业级软件开发的效率和质量。
在性能测试中，aiXcoder-7B Base表现卓越，不仅在HumanEval、MBPP和MultiPL-E三大主流评测集上的平均得分超越了拥有340亿参数的Codellama模型，更在代码补全场景下展现出同类参数量模型中最佳的效果。在多语言NL2Code基准测试中，aiXcoder-7B Base的平均表现也显著优于Codellama 34B和StarCoder2 15B。
这一开源项目不仅体现了北京大学在人工智能领域的深厚研发实力，也为全球企业提供了高效、安全的代码生成工具，有望推动软件开发行业的技术革新。aiXcoder-7B Base的开源，无疑将助力企业提升开发效率，降低出错率，进一步推动企业级软件开发的智能化进程。
英语如下：
**News Title:** “Peking University’s Open-Source Giant: aiXcoder-7B, Redefining Enterprise Code Development”
**Keywords:** aiXcoder-7B, Enterprise Deployment, Code Generation
**News Content:**
**Peking University’s aiXcoder team has recently launched the aiXcoder-7B Base, a groundbreaking large code model designed for enterprise private deployment, elevating the standards in software development.**
According to QbitAI, the aiXcoder team has once again made strides in the open-source domain with this new model. The aiXcoder-7B Base, with its exceptional performance and precision in code generation, ushers in a new era of breakthroughs in the software development industry.
Trained extensively on a massive 1.2T Unique Tokens dataset, the aiXcoder-7B Base aims to cater to real-world code generation scenarios. The model demonstrates unique ingenuity in pre-training tasks and context information processing, targeting enhanced efficiency and quality in enterprise-level software development.
In performance evaluations, the aiXcoder-7B Base outperformed Codellama, a model with 340 billion parameters, across the HumanEval, MBPP, and MultiPL-E benchmark tests. It also achieved the best performance among models with similar parameter counts in code completion scenarios. In the multilingual NL2Code benchmark, the aiXcoder-7B Base consistently surpassed Codellama 34B and StarCoder2 15B.
This open-source project not only underscores Peking University’s profound research capabilities in artificial intelligence but also offers global enterprises an efficient and secure tool for code generation. It has the potential to drive technological innovation in the software development industry. The open-source release of aiXcoder-7B Base is set to boost development efficiency, reduce error rates, and further propel the智能化进程 of enterprise software development.
【来源】https://www.qbitai.com/2024/04/134070.html
Views: 1