Browse Teach Curate

Computational Graphs

A directed acyclic graph whose nodes are operations and edges carry tensors; recording it during the forward pass is what lets automatic differentiation replay it backward for gradients.

jeremy Video

PyTorch Computational Graphs and Autograd Backward Pass Mechanics

Computational graphs in deep learning represent mathematical operations as a directed acyclic graph (DAG) that defines dependencies for automatic differentiation. The autograd mechanism leverages the…