Research and experimental code related to Opacus, a library that enables training PyTorch models with differential privacy.
Opacus Lab is meant to be an experimental counterpart of the main Opacus repository; it is used to include and experiment with features that are too niche or not mature to be included in the main repository.
⚠️ WARNING: This code is considered experimental and we provide it as-is to the community as we develop it in the open source. We do not offer support for this code, nor do we maintain a pip package for it with releases. You are welcome to install from source and tinker with it as well as talk to us about it, preferably in our forum (https://discuss.pytorch.org/c/opacus/29).
This code is released under Apache 2.0, as found in the LICENSE file.