Feb 06, 2026 Mixture of experts: why the largest LLMs use a few percent of their weights per token Jan 15, 2026 Hybrid attention: why new LLMs keep full attention in only a few layers May 07, 2025 LLM inference optimization: where the time and memory go Apr 17, 2025 Parameter-efficient fine-tuning in PyTorch: adapters, LoRA and the rest Mar 07, 2025 Building a Transformer in PyTorch, layer by layer