Cloudflare Cache Rules are zone-level cache policy: request expressions decide what is eligible for cache, how Edge TTL and Browser TTL behave, what dimensions enter the cache key, and how stale content, ETags, and purge interact. Use them for CDN cache policy; use the Worker Cache API for programmatic caching.
Cloudflare Images has two paths: transform images stored in R2/S3/origin at the edge, or store images in Images and deliver named variants. The first is priced by unique transformations; the second also involves stored and delivered images.
Smart Shield is Cloudflare's origin protection bundle: Smart Tiered Cache, connection reuse, Argo Smart Routing, Regional Tiered Cache, Cache Reserve, Health Checks, and Dedicated CDN Egress IPs reduce requests and connections reaching your origin.
Lecture 5 explains GPUs through SMs, warps, and the memory hierarchy, then unifies common optimization under low precision, fusion, recomputation, coalescing, and tiling. FlashAttention combines those principles for attention.
Lecture 6 turns GPU principles into kernels: benchmark scaling across shapes, profile actual calls and time, then implement GeLU, softmax, reductions, and tiled matrix multiplication in Triton. Speed begins with measuring correctly.
Lecture 2 reduces model training to tensors, FLOPs, bytes, and time: use einops to track dimensions, arithmetic intensity and roofline analysis to identify bottlenecks, then trade compute for memory with gradient accumulation and activation checkpointing.
CS107 Lecture 25 builds the essential cache model from a concise deck: memory access costs are nonuniform, smaller and faster layers retain data likely to be reused, and temporal and spatial locality determine whether a program benefits.
RAG system costs come from LLM tokens, Embedding APIs, and vector search. Every stage has room for cost reduction, but you need to verify that optimizations don't sacrifice too much quality.
Caching doesn't have to match exact query strings -- semantically similar questions can hit the cache too, skipping the entire RAG pipeline execution.