Intensive Program Teaches Acceleration of Large Language Models The LLM Scaling Week program, a collaboration between MIPT and Yandex Edu, focuses on educating participants on how to speed up large language models and optimize their usage effectively. The lectures, presented by YandexGPT team engineers, delve into crucial topics for practical applications, including communication structures in distributed training, modern techniques for enhancing LLM training efficiency (such as FP8, Triton, parallelisms, and Mixture of Experts), scaling from one to ten GPUs within budget constraints, and addressing deep learning arithmetic and inference challenges. This program is highly recommended for individuals aiming to improve the speed and cost-effectiveness of large language models, ultimately boosting their market competitiveness. To join the program, interested individuals can access the registration link provided.

The LLM Scaling Week program, a collaboration between MIPT and Yandex Edu, focuses on educating participants on how to speed up large language models and optimize their usage effectively.

The lectures, presented by YandexGPT team engineers, delve into crucial topics for practical applications, including communication structures in distributed training, modern techniques for enhancing LLM training efficiency (such as FP8, Triton, parallelisms, and Mixture of Experts), scaling from one to ten GPUs within budget constraints, and addressing deep learning arithmetic and inference challenges.

This program is highly recommended for individuals aiming to improve the speed and cost-effectiveness of large language models, ultimately boosting their market competitiveness. To join the program, interested individuals can access the registration link provided.