Skip to main content
T5 Others Sugon Round 2

Sugon AI Infra second round (2)

Real Sugon AI Infra second-round interview questions covering training and operator optimization

Training optimization Operator optimization

Basics

  1. Common PyTorch performance analysis tools and methods?
  2. How do you optimize DDP for multi-node multi-GPU?
  3. What should you watch when setting batch size in distributed training?
  4. How does PyTorch’s graph optimization work?
  5. What key new features did PyTorch 2.0 introduce?
  6. Common pytest parameters?
  7. How do you register operators from a YAML config in PyTorch?
  8. How should you design system-level operator tests across platforms?

Project experience

  1. Introduce core projects you worked on and their technical designs.