pytorch-summary

Why we included this project

Getting a clear picture of a PyTorch model from print(model) is a chore, and it tells you little about memory use. This utility prints a Keras-style summary instead: a layer-by-layer table with output shapes and parameter counts, along with estimates for input size and forward/backward pass size. Shape mismatches between modules become obvious at a glance, and the memory figures help you judge how large a batch your GPU can hold. The author has since pointed users to torchinfo as the maintained successor, so this repo is best treated as the lightweight original and a fast way to understand how the approach works. For older PyTorch code, or when you want one function and no extra dependencies, it still slots in easily.