Lablup wowed developers at PyCon Korea 2023, held August 11-13, with features and demos that automate the process of creating, fine-tuning, and servicing large language models with Backend.AI, its AI R&D and services platform software.

In the demo, various open-source large language models, including the recently released Llama2 model from Meta, were fine-tuned for Korean using Backend.AI's automation system, FastTrack. The chatbot, which was created on the fly with the Llama2 model fine-tuned in 12 hours, was serviced during the event via Lablup's NVIDIA H100 cloud farm with a dedicated UI, attracting the attention of many participants.

Backend.AI provides leading features in various AI development and service fields based on ▲ maximizing GPU cluster utilization, ▲ intuitive user experience and management, ▲ low-cost/low-latency AI model and service technology, and ▲ support for various AI semiconductors. With the unveiling of its large-scale language model development, fine-tuning, and service automation capabilities, Lablup has demonstrated that Backend.AI can be utilized for full-scale generative AI applications such as task automation, research, and programming code development.

"We have dramatically simplified the process of developing and integrating various services based on large language models into enterprises, institutions, research institutes, and organizations that require security. We are confident that any organization can use Backend.AI to easily fine-tune and operate their own unique large language model to meet their needs."

Currently in beta testing with select Backend.AI customers, the big language model integration will be available to on-premises customers as part of the general availability of Backend.AI in September. The service will also be on display at KSC2023 (August 21-23, The K Hotel Seoul), Korea's largest supercomputing conference and exhibition.

Automatically translated from Korean Article