tensorflow/privacy

Library for training machine learning models with privacy for training data

What it solves

TensorFlow Privacy helps developers train machine learning models while protecting the privacy of the data used to train them. It provides the tools necessary to implement differential privacy, a technique that ensures the model learns general patterns without memorizing specific details about individual data points.

How it works

The library provides specialized optimizers that you can wrap around standard TensorFlow optimizers like SGD or Adam. These optimizers perform "gradient clipping," which limits the influence of any single training example on the final model. The library also includes analysis tools to calculate and verify the privacy guarantees provided by your trained model.

Who it’s for

This library is designed for data scientists and machine learning engineers working with sensitive datasets who need to ensure their models comply with privacy standards. It is built to integrate directly into existing TensorFlow workflows.

Highlights

  • Includes specialized optimizers for training models with differential privacy.
  • Provides analysis tools to compute and measure privacy guarantees.
  • Features efficient per-example gradient clipping for Keras models, minimizing memory and runtime overhead.
  • Offers extensive tutorials to help users wrap existing optimizers and tune privacy parameters.
  • Compatible with TensorFlow 2.x and Keras models.

Related

  • Project
    tdrussell/diffusion-pipe

    diffusion‑pipe is an open‑source training framework that uses DeepSpeed pipeline‑parallelism to train huge diffusion models (image, text‑to‑image, text‑to‑video, etc.) across multiple GPUs. It provides caching of latents/text embeddings, TensorBoard logging, checkpoint/resume, and a unified code path for both image and video models. Install via Conda, set up PyTorch/Flash‑Attention, edit an example TOML config, and launch with `deepspeed`. The repo supports a long list of recent diffusion families (SDXL, Flux, Qwen‑Image, MiniMax H3, etc.) and includes low‑VRAM tricks for training 14‑B‑parameter models on a single GPU.

  • Project
    Azure/azureml-examples

    A collection of examples and tutorials for Azure Machine Learning, covering Python, .NET, TypeScript SDKs and the Azure CLI extension, with a focus on the Python SDK v2.

  • Project
    tensorflow/recommenders

    A TensorFlow-based library for building recommender system models, supporting the full workflow from data preparation to deployment.

  • Project
    pykeen/pykeen

    PyKEEN is a Python library for training and evaluating knowledge‑graph embedding models, offering dozens of models, many benchmark datasets, and a high‑level pipeline API.