TAIPEI, Taiwan, March 8, 2023 — ASUS today announced that Taiwan Web Services (TWS) has launched its AI 2.0 Foundation Model Consulting Services — a one-stop solution that integrates infrastructure, development environment, and professional technical team services for the development of next-generation AI. TWS is the first company in Taiwan to integrate a BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) into a supercomputer.
Despite the recent rapid development of large language models (LLM) and generative AI, it is still tough for an enterprise to conduct an LLM project by itself. That is why using the TWS one-stop AI 2.0 Foundation Model Consulting Services is so powerful — it dramatically reduces the barriers to entry and allows enterprises to concentrate more on research and development projects.
Using the new TWS service, enterprises can immediately start building their own generative AI applications while simultaneously reducing their hardware equipment and human capital costs, as well as lowering development risk and time to completion.
Lower barriers to entry for LLM projects
AI technology combined with large language modeling has become an important trend in industrial technology development. But, it is still challenging for an enterprise to conduct an LLM project by itself. That is why using the TWS one-stop AI 2.0 Foundation Model Consulting Services is so powerful — it dramatically reduces the barriers to entry and allows enterprises to concentrate more on research and development projects.
“The TWS AI 2.0 platform, and its technical resources team, provide complete external support for the building of AI applications,” Steve Chen, CTO of TWS, explained. "The new service model is already attracting a range of companies in the high-tech R&D, manufacturing, financial, and retail sectors.”
Innovation and performance
The BLOOM’s data set contains 46 human languages and 13 programming languages, with 176 billion parameters, adding up to more than 1.5 TB of data. Additionally, 840 GPUs were used for node training, which approximates the linear theoretical value and converges the training results. If users want to increase the performance, they can also scale out linearly and increase the GPU to improve parallel computing capability.
However, the deployment threshold for LLM projects is extremely high as enterprises need to be familiar with how to build and fine-tune an Artificial Intelligence High-Performance Computing (AIHPC) system as well as how to conduct distributed training and large model inference.
But this is not the case if an enterprise uses the AI 2.0 Foundation Model Consulting Services to create their own generative AI applications. They will be able to quickly reduce costs on things like hardware equipment and human capital, as well as lowering development risk and time to completion.
TWS recently showcased its AIHPC capabilities in the research and development of foundation models at Supercomputing Asia 2023 (SCA23) in Singapore. TWS Vice President of Enterprise & Ecosystem Development Division, Andy Huang, introduced TWS AI 2.0 as one of the highlights of the event, showcasing the innovative and comprehensive one-stop service, which received widespread attention and positive feedback.
TWS and AIHPC supercomputing
TWS is a subsidiary of ASUS that provides a world-class AI supercomputer cloud platform. Its Taiwania 2 is Asia’s first commercial AIHPC supercomputer that caters to the service industry. With a computing power of nine PFLOPS, it balances high efficiency with user-friendly accessibility and Environmental, Social, and Governance (ESG) considerations, helping businesses with digital transformation and green initiatives. It is an indispensable commercial partner in the AI 2.0 era.
For more information on the TWS AI 2.0 Foundation Model Consulting Services, please go to:
https://tws.twcc.ai/en/ai-llm/