OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Efficient Densenet Pytorch

A Python implementation of DenseNet for PyTorch that uses activation checkpointing to reduce GPU feature-map memory during training.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Efficient Densenet Pytorch — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use efficient=True when GPU memory is the constraint and the documented training-time overhead is acceptable; use efficient=False when memory is not a concern and speed is the priority.

Who it is for

  • PyTorch developers integrating DenseNet into an existing project.
  • Researchers or practitioners training DenseNet models where GPU memory use is a constraint.
  • People configuring DenseNet experiments for CIFAR, SVHN, or ImageNet inputs.

Documented uses

  • Use a checkpointing-based DenseNet implementation in a PyTorch project.
  • Trade additional training time for reduced feature-map memory use.
  • Configure model depth, growth rate, and input settings for an experiment.
Before you choose

Requirements & limits

The repository identifies the project as MIT-licensed open-source software.

  • Efficient mode recomputes intermediate batch-normalization and concatenation feature maps during backpropagation.
  • The README states a 15–20% training-time overhead for checkpointing.
  • The documented requirements are PyTorch 1.0 or later and CUDA.
  • The README provides separate small-input settings for CIFAR or SVHN and ImageNet.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Configure the DenseNet implementation for a project

Choose the documented memory mode and input setting when incorporating the model.

Before you start

  • PyTorch 1.0 or later
  • CUDA
  • The implementation's model file in the repository's models folder

Steps

  1. Use the DenseNet implementation from the models folder in the PyTorch project.

  2. Set efficient=True for checkpointing-based memory savings, or efficient=False when prioritizing speed over memory.

  3. Set small_inputs=True for CIFAR or SVHN, or small_inputs=False for ImageNet.

  4. Adjust block_config for network depth and growth_rate for features added per DenseNet layer.

Expected outcome

A DenseNet configuration with the documented memory mode, input setting, depth, and growth rate.

Guide limitations

  • Efficient mode adds a stated 15–20% training-time overhead.
  • This guide covers model configuration, not execution or training verification.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
gpleiss/efficient_densenet_pytorch · Official homepage title · retained in its source language
How we review sourcesSuggest a correction