quark0/darts
A PyTorch-based differentiable architecture search algorithm that automatically discovers convolutional and recurrent network architectures via gradient descent.
Not currently ranked — collecting fresh signals.
star history
DARTS implements continuous relaxation to make neural architecture search differentiable, enabling efficient gradient-based optimization over the architecture space. The algorithm searches for optimal operations and connections in CNNs for image classification and RNNs for language modeling, requiring only a single GPU. It produces architectures competitive with manually designed networks on CIFAR-10, ImageNet, Penn Treebank, and WikiText-2 benchmarks.
Frequently asked
- What is quark0/darts?
- A PyTorch-based differentiable architecture search algorithm that automatically discovers convolutional and recurrent network architectures via gradient descent.
- Is darts open source?
- Yes — quark0/darts is open source, released under the Apache-2.0 license.
- What language is darts written in?
- quark0/darts is primarily written in Python.
- How popular is darts?
- quark0/darts has 4k stars on GitHub.
- Where can I find darts?
- quark0/darts is on GitHub at https://github.com/quark0/darts.