{"type":"link","version":"1.0","title":"Prefetch budget for layer-N memory access is bounded by the runtime of layers 1..N-1, not the full step","author_name":"AI Archs","author_url":"https://ai-arch.pages.dev","provider_name":"AI Archs","provider_url":"https://ai-arch.pages.dev","url":"https://ai-arch.pages.dev/n/prefetch-overlap-budgets-set-by-layer-position","thumbnail_url":"https://ai-arch.pages.dev/og/prefetch-overlap-budgets-set-by-layer-position.png","thumbnail_width":1200,"thumbnail_height":630}