The recent rapid development of Large Language Models (LLMs) and Generative AI has ushered in a new era of AI. Taizhi Cloud has taken the lead in Taiwan by announcing that it has become the first company in Taiwan to successfully build a BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) model with 176 billion parameters. Taizhi Cloud has launched a one-stop integrated solution, "AI 2.0 High-Computing Power Consulting Service," which integrates the infrastructure, development environment, and professional technical team services for next-generation AI. Enterprises can immediately invest in AI application development and enter the AI 2.0 era through Taizhi Cloud's services.
Chen Zhongcheng, Chief Technology Officer of Taizhiyun, pointed out: "The combination of AI technology and large-scale language models is already a trend in industrial technology development. It is extremely difficult for enterprises to execute LLM projects independently. By utilizing the one-stop AI 2.0 high-computing power consulting service, they can focus more on R&D projects and significantly lower the entry barrier. Taizhiyun's AIHPC high-speed computing platform, combined with its technical resource team, provides complete external AI support with a brand-new service model. Currently, companies in the high-tech R&D and manufacturing, finance, and retail sectors are in talks for cooperation."
Taking the BLOOM large-scale model completed by Taizhiyun as an example, the dataset includes 46 human languages and 13 programming languages, with 176 billion parameters and a total data volume of over 1.5TB. It utilizes 840 GPUs for cross-node training. In addition to achieving performance close to the theoretical linear value, it can also converge the training results. If further performance is desired, it can be scaled out linearly by adding GPUs to enhance parallel computing capabilities. Furthermore, enterprises that promote LLM projects on their own need to be familiar with distributed training architectures, build AIHPC high-speed computing systems, understand fine-tuning techniques, and large-scale model inference. The deployment threshold is extremely high. However, by using the "AI 2.0 High-Computing Power Consulting Service," the costs of time investment, technical costs, development risks, hardware equipment, and human resources investment can be quickly reduced, enabling the creation of proprietary generative AI applications.