智源研究院正式开源了700亿参数的大模型Aquila2-70B-Expr，这是首个基于英伟达混合资源（A100集群+A800集群）和天数智芯混合资源（BI-V100集群+BI-V150集群）完成训练的大模型。Aquila2-70B-Expr的训练使用了异构先驱版，并且采用了混合并行训练的方法，从而实现了对数据的高效处理。该模型的开源举措旨在促进人工智能的研究和发展，为智源研究院的研究人员提供一个强大的工具。
新闻翻译：
Title: First open-source Aquila2-70B-Expr large model, trained on mixed resources of Nvidia and Tianchi
Keywords: Aquila2-70B-Expr, large model, open-source, Nvidia, Tianchi
News content:
Zhiyuan Research Institute has officially opened up the large model Aquila2-70B-Expr, which is the first large model to be trained using mixed resources of Nvidia and Tianchi. The training of Aquila2-70B-Expr uses the heterogeneous pioneer version and adopts the mixed parallel training method, which enables efficient data processing. The open-source initiative aims to promote the research and development of artificial intelligence and provides a powerful tool for the researchers at Zhiyuan Research Institute.
【来源】https://www.chinastarmarket.cn/detail/1531570
Views: 1