Search Articles

Discover what Large Language Models (LLMs) are and how they power Generative AI. Learn about pre-training, fine-tuning, and the Transformer architecture.

Explore LLM inference optimization on H800 SuperPods. Learn how a disaggregated architecture with SGLang tackles the prefill bottleneck to boost throughput.

This article delves into the core challenges of GPU performance, analyzing the differences between compute-bound and memory-bound operations and highlighting the issue of underutilized memory bandwidth. It further proposes strategies to maximize throughput and looks ahead to the collaborative future of CPUs and GPUs, as well as the evolution of GPU architecture, offering a first-principles perspective on understanding and optimizing GPU performance.

This article dissects the architectural evolution of modern large language models in 2025, moving beyond benchmarks to analyze the core design choices of flagship open-source models. We explore key innovations like DeepSeek-V3's Multi-Head Latent Attention (MLA) and Mixture of Experts (MoE), OLMo 2's unique normalization strategies, Gemma 3's use of sliding window attention, and Llama 4's take on MoE. By focusing on these architectural blueprints, we gain a clearer understanding of the engineering priorities shaping the future of LLMs.

Learn how Supervised Fine-Tuning (SFT) transforms LLMs from base models to chat assistants. Step-by-step guide to SFT workflow, datasets, and best practices.

Discover how linear layers enable multi-head attention in Transformers, powering advanced NLP models with parallel processing and rich representations.

Discover the technical challenges and solutions in training a 671B parameter LLM with Reinforcement Learning, covering frameworks, memory, and efficiency.