to leave a comment.
SK Telecom [017670] announced on the 22nd that it has disclosed cases of cooperation with Nvidia in AI model development.
SKT announced its AI collaboration plans, including the development of A.X K2 (A.X K2), at Nvidia's Nemotron Developer Days held on the 21st.
The two companies have been collaborating across data, infrastructure, and learning environments, starting from the construction of the A100 GPU-based supercomputer Titan, which SKT unveiled in 2021.
In particular, SKT adopted Nvidia's Nemotron dataset for model training while carrying out the government's independent AI foundation model project last year.
In this process, the two companies enhanced the stability of AI training by utilizing Megatron LM, a large-scale model distributed training framework, and Nvidia Nemotron Curator.
The two companies shared their experience in creating hyperscale models with new architectures like MoE (Mixture of Experts) and information related to the development infrastructure supporting them, establishing a foundation for stable model training.
SKT plans to continue its collaboration with Nvidia and utilize Nvidia solutions for the training of A.X K2, which is currently under development.
Newsletter
Get key news delivered to your email every morning
to leave a comment.