GitLab

PiperOrigin-RevId: 394327712

PiperOrigin-RevId: 394303266

PiperOrigin-RevId: 394299699

This also concretizes some generic methods used in previous model serialization
that appeared to interact poorly with tf.function tracing.

PiperOrigin-RevId: 394295569

* Added BUILD files.
* Updated WORKSPACE.
* Updated bazelrc.
* Added TF patch.

PiperOrigin-RevId: 394259252

Adding implementations of private heavy hitters with invertible Invertible Bloom Lookup Tables (IBLT).

PiperOrigin-RevId: 394247532

Add DpEvent to return value of get_noised_result. For most DPQueries, the default UnsupportedDpEvent is returned, pending further development.

PiperOrigin-RevId: 394137614

See https://github.com/grpc/grpc/pull/13929 for more information.

PiperOrigin-RevId: 394112774

PiperOrigin-RevId: 394111901

PiperOrigin-RevId: 394099367

PiperOrigin-RevId: 394089953

PiperOrigin-RevId: 393891207

PiperOrigin-RevId: 393880481

Create a library containing client and server measurements suitable for debugging learning processes, and are intended for use with tff.aggregators.add_measurements.

PiperOrigin-RevId: 393843919

This will allow for serializable Model that are also composable post
deserialization, as well as more Jax-friendly APIs to go along with
tensorflow_federated/experimental/python/learning/jax_components.py

PiperOrigin-RevId: 393835777

PiperOrigin-RevId: 393815864

PiperOrigin-RevId: 393790129

PiperOrigin-RevId: 393174542

PiperOrigin-RevId: 392999200

Fixes #1747

PiperOrigin-RevId: 392960630

Create a composer for computations mirroring the structure of federated evaluation. This change also creates re-usable layers for client work and metrics aggregation.

PiperOrigin-RevId: 392952848

PiperOrigin-RevId: 392931812

PiperOrigin-RevId: 392785924

PiperOrigin-RevId: 392762446

PiperOrigin-RevId: 392758851

Sets noise_multiplier and clients_per_round correctly in construction of DifferentiallyPrivateFactory in aggregator tuning colab.

PiperOrigin-RevId: 392687645

PiperOrigin-RevId: 392470163

PiperOrigin-RevId: 392452227

PiperOrigin-RevId: 392084010

PiperOrigin-RevId: 392078655

PiperOrigin-RevId: 392073116

PiperOrigin-RevId: 391889931

PiperOrigin-RevId: 391879106

Logarithm of timestamp will return 35 for next many years thus not suitable for random seed generation :|

PiperOrigin-RevId: 391689896

This resolves a bug where the C++ will fail if a `tff.Computation` is returning
multiple variant tensors from the same node (e.g. a `PartitionedCall` op).

PiperOrigin-RevId: 391638014

This triggers a bug in the C++ runtime that fails to uniquely name the DatasetToGraphV2 JIT nodes.

PiperOrigin-RevId: 391626904

Upstream tree aggregation factory to tff.aggregators. The aggregator can give us differential privacy without relying on sampling for amplification. More details see "Practical and Private (Deep) Learning without Sampling or Shuffling" (https://arxiv.org/abs/2103.00039).

PiperOrigin-RevId: 391409199

PiperOrigin-RevId: 391269837

Remove `@tf.function` decorator for `predict_on_batch` to workaround a tf.data performance regression.

PiperOrigin-RevId: 390692772

Add a call to `tf.convert_to_tensor` within `tff.simulation.datasets.TestClientData` in order to allow passing strings to `.serializable_dataset_fn`.

This change also adds tests that will fail without this conversion to tensor.

PiperOrigin-RevId: 390674662