# gpleiss/efficient_densenet_pytorch

A memory-efficient implementation of DenseNets

Repository: https://github.com/gpleiss/efficient_densenet_pytorch
Canonical: https://ross.abutalabs.com/products/efficient_densenet_pytorch
Language: Python
License: MIT
License Family: permissive
Topics: densenet, pytorch, deep-learning
Last push: 2023-06-01T13:11:51+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3381, "days_push": 1189, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1535, forks 320 (observed 2026-08-28T04:04:59.797918+00:00)

## What it is
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
- artifact type: library
- maturity: maintenance
- function: deep-learning, machine-learning, image-processing
- domain: deep-learning, machine-learning, computer-vision
- platform: python
- tags: densenet, pytorch, memory-efficient, checkpointing, gpu-memory, image-classification, gpu

## Member repositories
- gpleiss/efficient_densenet_pytorch (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:59.797918+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:31:09.107000+00:00, confidence not recorded.
  - readme: https://github.com/gpleiss/efficient_densenet_pytorch (fetched 2026-08-28T04:04:59.797918+00:00, sha 95b4eb5de461)
- Data as of 2026-08-30T08:39:29.467469+00:00.