Level 12: Optimize training speed

In this level you'll use compilers, advanced profilers and mixed precision techniques to train bigger models faster.

• Speed up models by compiling them — Use torch.compile to speed up models on modern hardware

• Explore advanced mixed precision settings — Enable state-of-the-art scaling with advanced mix-precision settings.

• Enable advanced profilers — Tune model performance with profilers.

• Profile PyTorch operations — Learn to find bottlenecks in PyTorch operations.