跳转至

Lecture 2: PyTorch & Resource Accounting

约 51 个字 预计阅读时间不到 1 分钟

Abstract

Overview of this Lecture:

  • Discuss all the premitives needed to train a model;
  • Go bottom-up from tensors to models to optimizers to the training loop;
  • Pay close attention to efficiency.

Takeaways:

  • Mechanics: straightforward (PyTorch);
  • Mindset: resource accounting (remember to do it).

Memory Accounting

Compute Accounting