nemo_microservices.types.beta.safe_synthesizer.differential_privacy_hyperparams_param#
Module Contents#
Classes#
API#
- class nemo_microservices.types.beta.safe_synthesizer.differential_privacy_hyperparams_param.DifferentialPrivacyHyperparamsParam#
Bases:
typing_extensions.TypedDict- delta: typing_extensions.Literal[auto] | float#
None
Probability of accidentally leaking information.
Setting to ‘auto’ usesdelta of 1/n^1.2, where n is the number of training records
- dp_enabled: bool#
None
Enable differentially-private training with DP-SGD.
- epsilon: float#
None
Target for epsilon when training completes.
- per_sample_max_grad_norm: float#
None
Maximum L2 norm of per sample gradients.