Generative Models trained with federated learning using TensorFlow Federated.
Code was executed using Python 3.8
Please run pip install -r requirements.txt
We use hydra experiment configuration.
The config files can be found in the ./config directory.
They are split thematically and define the experiment structure
- Hydra mutliruns (
python main.py -m ...) don't work without defining more sophisticated launchers (such ashydra-submitit). The issue is that@tf.functiondecorated functions are already instantiated for the 2nd and further runs which throws an exception.