【北京大学开源创新:aiXcoder-7B 代码大模型引领企业软件开发新纪元】

近日,北京大学的aiXcoder团队在人工智能领域再次取得重大突破,开源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,为软件开发行业带来了全新的解决方案。

据悉,aiXcoder-7B Base版在1.2T Unique Tokens的海量数据上进行了深度训练,其设计独特,预训练任务及上下文信息均针对实际的代码生成场景进行了优化。在HumanEval、MBPP和MultiPL-E三大主流代码生成评测集上,该模型的平均得分超越了拥有340亿参数的Codellama,显示出卓越的代码补全能力。

在多语言自然语言到代码转换(NL2Code)基准测试中,aiXcoder 7B Base的表现同样出色,平均效果优于Codellama 34B和StarCoder2 15B等同级别参数量的模型。这一成就标志着北京大学在人工智能领域的领先地位,以及在代码生成技术上的显著进步。

aiXcoder-7B Base的开源,将有望降低企业软件开发的复杂性和成本,提升开发效率,同时也为全球的开发者提供了一个高效、智能的工具,有望引领新一轮的企业级软件开发创新热潮。北京大学的这一创新举措,再次彰显了中国在人工智能科研领域的强劲实力和开放共享的精神。

英语如下:

**News Title:** “Peking University Sets a New Benchmark in Enterprise Code Generation with Open-Source aiXcoder-7B”

**Keywords:** aiXcoder-7B, enterprise deployment, code generation

**News Content:**

**[Peking University Leads Open-Source Innovation: aiXcoder-7B, A New Era in Enterprise Software Development]**

Recently, the aiXcoder team at Peking University has made a groundbreaking advancement in artificial intelligence by open-sourcing aiXcoder-7B Base, a large code generation model specifically designed for enterprise private deployment. This model, with its impressive performance and precision in code generation, introduces a revolutionary solution to the software development industry.

aiXcoder-7B Base has been extensively trained on a massive dataset of 1.2 trillion unique tokens. Its unique design and pre-training tasks, optimized for real-world code generation scenarios, set it apart. Outperforming Codellama, a model with 340 billion parameters, on the HumanEval, MBPP, and MultiPL-E benchmark code generation tasks, aiXcoder-7B demonstrates exceptional code completion capabilities.

In the NL2Code benchmark for natural language to code translation, aiXcoder 7B Base also excels, surpassing models like Codellama 34B and StarCoder2 15B with comparable parameter counts. This milestone underscores Peking University’s leading position in AI research and its significant strides in code generation technology.

The open-source release of aiXcoder-7B Base is anticipated to reduce complexity and costs in enterprise software development, boosting efficiency. It also offers a powerful and intelligent tool to developers worldwide, potentially sparking a new wave of innovation in enterprise-level software development. Peking University’s innovative move underscores China’s robust capabilities in AI research and its commitment to openness and collaboration.

【来源】https://www.qbitai.com/2024/04/134070.html

Views: 1

发表回复

您的电子邮箱地址不会被公开。 必填项已用 * 标注