Updates to Opacus to handle extra args returned in ghost clipping, dynamically provide epsilon_tolerance. #695
Conversation
…rithm is described in the 'A Unified Fast Gradient Clipping Framework for DP-SGD' paper: https://proceedings.neurips.cc/paper_files/paper/2023/file/a45d344b28179c8da7646bc38ff50ad8-Paper-Conference.pdf. This reduces the memory needed to run DP-SGD over embedding layers, significantly reducing OOMs over large embedding layers.
… returned for ghost vs non-ghost clipping techniques. The approach can be further enhanced, adding a quick fix to handle the error. b) Allow providing epsilon_tolerance dynamically.
|
Hi @pagarwl! Thank you for your pull request and welcome to our community. Action RequiredIn order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at [email protected]. Thanks! |
facebook-github-bot
added
the
CLA Signed
|
Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks! |
|
@facebook-github-bot has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
Types of changes
Motivation and Context / Related issue
a) Currently make_private returns
criterionif the ghost clipping algorithm was chosen, otherwise it doesn't. The return args listTuple[GradSampleModule, DPOptimizer, DataLoader]is incorrect in this case. This causes compile time error when using ghost clipping technique. Removing the return arg specification (similar to make_private_with_epsilon), as a quick fix to avoid compile errors, while the approach can be further updated.b) Allow providing epsilon_tolerance dynamically.
How Has This Been Tested (if it applies)
Compile, unit tests, and manual runs.
Checklist