Faster Sampling And Training Optimization Techniques

Python materials for the online course on diffusion models by [@huggingface](https://github.com/huggingface).

Unique: Provides systematic comparison of sampling schedulers (DDIM, Euler, DPM++) with explicit speed/quality tradeoffs, and covers training optimizations including gradient checkpointing and xFormers attention. The course demonstrates how to measure actual speedups and validate that optimizations don't degrade output quality.

vs others: More practical than benchmark papers because it includes code examples and tuning guidance; more comprehensive than single-optimization tutorials because it covers both inference and training acceleration.