PiperOrigin-RevId: 271416217

PiperOrigin-RevId: 268787966

PiperOrigin-RevId: 268493485

PiperOrigin-RevId: 266383040

PiperOrigin-RevId: 259669627

This enables the tools from tfmot(.)tensor_encoding package to be used to specify encoding mechanism in the StatefulBroadcastFn. The enclosed next_fn first encodes the value on the tff.SERVER using provided Encoder, broadcasts the encoded representation, and decodes on tff.CLIENTS.

PiperOrigin-RevId: 258310416

PiperOrigin-RevId: 249716137

PiperOrigin-RevId: 249713486

- Remove TensorFlow repository in Workspace
- Add Protobuf, zlib, and six dependencies that were previously coming from TF
- Add version match strings to requirements.txt
- This requires the environment to provide the dependent packages.

PiperOrigin-RevId: 245328284

PiperOrigin-RevId: 242473485

PiperOrigin-RevId: 239708605

The current default works well for federated_averaging, but is not necessairly
generalizable to all federated optimizations.

PiperOrigin-RevId: 236383914

Preserve the order of variable lists and provide ModelWeights.keras_weights, and add methods to convert (keras model weights) <--> (tff model state).

This lets us get back from a ModelWeights structure to a list that can be used with keras_model.set_weights().

Fixes a bug where anonymous_tuple.flatten was used together with nest.pack_sequence_as, which happened to work before but didn't after the other changes in this CL. Consolidates all the TFF-specific code in optimizer_utils.py to a single function.

Raise a ValueError if someone tries to construct an AnonymousTuple with duplicated names, and also check to this when construct OrderedDicts of variables.

PiperOrigin-RevId: 234265469

Keeps shape information for Datasets constructed from empty elements and type_spec, and guards against nan in server model at federated_averaging level.

PiperOrigin-RevId: 234233165

PiperOrigin-RevId: 233541246

PiperOrigin-RevId: 231443201

- Add the "federated orchestration" logic to tie together local client training with global averaging and global model updates.
- Rename build_model_delta_optimizer_tff -> build_model_delta_optimizer_process
- Add a new test utility class that matches any type.

PiperOrigin-RevId: 230965510

PiperOrigin-RevId: 230803223

* Moved test functions into one location.
* Renamed test functions and module to be more concisely named.
* Moved TF 2.0 setup from setup methods to main function.

PiperOrigin-RevId: 230776467

PiperOrigin-RevId: 228361902

- Replace SequentialTffComputation with tff.IteratedProcess, avoid confusion with the tff.Computation type
- Rename federated_averaging() -> build_federated_averaging_process() to adhere to 'methods have verb-like naming' convention

PiperOrigin-RevId: 227941086

* Added missing BUILD dependencies.
* Added six python dependencies.
* Fixed whitespace.

PiperOrigin-RevId: 227725720

Adds initial TFF orchestration for optimizers based on computing model deltas (FedAvg and FedSgd). Also moves common server optimizer functionality to framework/optimizer_utils.py, and adds tests. Moves @graph_mode_test to test_utils.py

PiperOrigin-RevId: 227157965

Renaming to clarify the relationship between trainable, non-trainable, and local variables. Also resolves the issue that 'model' and 'initial_model' were different types, which was surprising. In particular:
  * initial_model => initial_weights
  * model_delta => weights_delta
  * ModelVars => ModelWeights
  * enhanced_model.vars => enhanced_model.weights

Also regenerates documentation.

PiperOrigin-RevId: 227040535

Implements the client logic for federated_sgd, and adds a common interface for client optimizers that average model deltas (FedSgd and FedAvg).  Common functionality for building federated optimizers moves to framework/optimizer_utils.py.

PiperOrigin-RevId: 227026404

Explicitly factors out a 'tff.learning.framework' namespace that will constitute the API surface for contributors to federated learning infrastructure.

PiperOrigin-RevId: 225904387