《动手学深度学习》：面向中文读者、能运行、可讨论。中英文版被70多个国家的500多所大学用于教学。

xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism

A simple python script to generate dot files for vizualization of call graph from function call traces with trace-cmd command.