gpleiss/efficient_densenet_pytorch
A memory-efficient PyTorch implementation of DenseNet convolutional neural networks for image classification.
Not currently ranked — collecting fresh signals.
star history
This repository provides a PyTorch implementation of DenseNets that reduces GPU memory consumption using gradient checkpointing. Rather than storing intermediate feature maps during the forward pass, the implementation recomputes them during backpropagation. This trades 15-20% extra compute time for linear instead of quadratic memory scaling with network depth. The implementation is based on torchvision’s DenseNet with options to toggle between memory-efficient and speed-optimized modes.
Frequently asked
- What is gpleiss/efficient_densenet_pytorch?
- A memory-efficient PyTorch implementation of DenseNet convolutional neural networks for image classification.
- Is efficient_densenet_pytorch open source?
- Yes — gpleiss/efficient_densenet_pytorch is open source, released under the MIT license.
- What language is efficient_densenet_pytorch written in?
- gpleiss/efficient_densenet_pytorch is primarily written in Python.
- How popular is efficient_densenet_pytorch?
- gpleiss/efficient_densenet_pytorch has 1.5k stars on GitHub.
- Where can I find efficient_densenet_pytorch?
- gpleiss/efficient_densenet_pytorch is on GitHub at https://github.com/gpleiss/efficient_densenet_pytorch.