facebookresearch/higher
A PyTorch library providing higher-order gradient support and differentiable optimizers for meta-learning algorithms.
Not currently ranked — collecting fresh signals.
star history
Higher is a library enabling higher-order optimization over training loops, specifically for meta-learning applications. It makes torch.nn.Module instances stateless so parameter changes can be tracked and gradients taken through intermediate parameters. The library provides differentiable optimizers to facilitate implementing meta-learning approaches like MAML (Model-Agnostic Meta-Learning), allowing computation of gradients with respect to initial parameters through multiple optimization steps.
Frequently asked
- What is facebookresearch/higher?
- A PyTorch library providing higher-order gradient support and differentiable optimizers for meta-learning algorithms.
- Is higher open source?
- Yes — facebookresearch/higher is open source, released under the Apache-2.0 license.
- What language is higher written in?
- facebookresearch/higher is primarily written in Python.
- How popular is higher?
- facebookresearch/higher has 1.6k stars on GitHub.
- Where can I find higher?
- facebookresearch/higher is on GitHub at https://github.com/facebookresearch/higher.