search

Trusted-AI / adversarial-robustness-toolbox

Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue Teams
https://adversarial-robustness-toolbox.readthedocs.io/en/latest/
MIT License
4.77k stars 1.15k forks source link

Freeze Model Weights #1451

Open keykholt opened 2 years ago

keykholt commented 2 years ago

Is your feature request related to a problem? Please describe. Fine tuning models requires the ability to free model weights. ART's estimators does not seem to expose this functionality as part of the default API. Currently, freezing model weights needs to be done using external methods depending on the underlying framework.

Describe the solution you'd like A freeze_layers() method that is part of the estimator API, which can freeze all or a subset of the layers depending on the input arguments.

keykholt commented 2 years ago

I can work on implementing this for each of the different frameworks

beat-buesser commented 2 years ago

Hi @keykholt Thank you very much! I think this will be a very useful feature!