gpleiss/efficient_densenet_pytorch

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1535 stars · 320 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A memory-efficient PyTorch implementation of DenseNets that uses gradient checkpointing to reduce feature map memory consumption from quadratic to linear in network depth. It trades roughly 15-20% training time overhead for significantly lower GPU memory usage.

Use cases

• train DenseNets on GPUs with limited memory
• fit larger batch sizes when training DenseNet image classifiers
• train deep DenseNets on CIFAR-10 or SVHN without running out of GPU memory
• swap a memory-hungry DenseNet implementation for an efficient one in an existing PyTorch project
• train DenseNet on ImageNet with multi-GPU setups

When to choose

• you need to train DenseNets but hit GPU out-of-memory errors
• you want larger batch sizes or deeper DenseNet models within fixed GPU memory
• you can tolerate a small training time overhead in exchange for linear memory usage

When to avoid

• training speed is your top priority and memory is not a constraint
• you need a general-purpose model zoo rather than DenseNet specifically
• you require actively maintained code with frequent updates, as the project is in maintenance mode

Facets

library · maturity maintenance

deep-learning machine-learning image-processing deep-learning machine-learning computer-vision python densenet pytorch memory-efficient checkpointing gpu-memory image-classification gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="gpleiss/efficient_densenet_pytorch")