An open-source framework for machine learning and other computations on decentralized data.
APACHE-2.0 License
Bot releases are hidden (Show)
Published by michaelreneer over 2 years ago
tff.program
API; this API is still in active development but can betff.simulation.datasets
.tff.analytics
API for federated analytics, including private heavytff.learning.algorithms
API, including TFF implementations oftff.learning.metrics
API to support easy configuration ofmetrics_aggregator
argument.metrics_aggregator
argument totff.learning.build_federated_averaging_process
andtff.learning.build_federated_evaluation
.report_local_unfinalized_metrics
and metric_finalizers
methods totff.learning.Model
and deprecated report_local_outputs
andfederated_output_computation
.tff.learning.optimizers
API for building purely functionaltff.learning.reconstruction
API for building partially localtff.learning.templates
API to support building learning algorithmstff.simulation.baselines
API to support evaluating learningtff.aggregators.DifferentiallyPrivateFactory.tree_aggregation
totff.aggrgators.SecureModularSumFactory
tff.aggregators.DiscreteFourierTransformFactory
andtff.aggregators.HadamardTransformFactory
to support rotation-basedtff.aggregators.concat_factory
for aggregating structures as atff.backends.native.create_mergeable_comp_execution_context
,tff.backends.native.set_mergeable_comp_execution_context
; these can betff.simulation.datasets.ClientData
tff.simulation.datasets.ClientData.serializable_dataset_fn
debug_measurements
option to aggregators in tff.learning
.tff.federated_zip
to invocation oftff.simulation.datasets
for saving federated datasetstff.simulation.datasets.SqlClientData
.tff.learning.models.FunctionalModel
andtff.learning.models.functional_model_from_keras
.absl-py
to version 1.0.0
.attrs
to version 21.2.0
.farmhashpy
version 0.4.0
.jax
to version 0.2.27
.jaxlib
to version 0.1.76
.numpy
to version 1.21.4
.retrying
.tensorflow-model-optimization
to version 0.7.1
.tensorflow-model-optimization
to version 0.7.3
.tensorflow
to version 2.8.0
.tensorflow
usingrules_python
to version 0.5.0
.com_google_protobuf
to version v3.18.0-rc1
.absl_py
version 1.0.0
.com_google_googletest
version release-1.11.0
.io_bazel_rules_go
version v0.29.0
.bazel_skylib
version 1.0.3
.pybind11_abseil
.pybind11_bazel
.pybind11_protobuf
.com_google_absl
version 20211102.0
.tensorflow_org
version v2.8.0
.tff.framework.type_contains
, use tff.types.contains
tff.simulation
, these can be found intff.program
instead.tff.Value
s.tff.simulation.datasets.ClientData.from_clients_and_fn
in favor oftff.simulation.datasets.ClientData.from_clients_and_tf_fn
.tff.simulation.datasets.ClientData.preprocess
to only supporttff.backends.reference
, and the reference context it contained.tff.learning.build_federated_sgd_process
in favor oftff.learning.algorithms.build_fed_sgd
.tff.simulation.run_simulation
in favor oftff.simulation.run_training_process
.tff.learning.framework.EnhancedModel
.tff.learning.framework.build_stateless_mean
.tff.simulation.datasets.gldv2.load_data()
would result inPublished by ZacharyGarrett over 3 years ago
federated_select
and federated_secure_select
.tff.structure_from_tensor_type_tree
to help manipulate structures oftff.TensorType
into structures of values.tff.aggregators
factory implementations.tf.data
concept for data URIs.tff.type
package with utilities for working with tff.Type
values.tff.jax_computation
.tff.tf_computation
support to SpareTensor
and RaggedTensor
.ClientData
interface and implementations totff.simulation.datasets
.tff.utils.update_state
to tff.structure.update_struct
.tff.utils
namespace, all symbols have migrated, many totff.aggregators
.rpc_mode
argument to remote executors, along with streaming mode.tff.federated_apply
.tff.federated_reduce
, all usages can usetff.federated_aggregate
.HDF5ClientData
and h5py
pip dependency.setattr
functionality on tff.ValueImpl
.tf.GraphDef
comparisons.abc
.ClientData.create_dataset_from_all_clients
consuming unreasonablePublished by michaelreneer over 3 years ago
tff.simulation
package to add many new tools for runningtff.aggregators
package with a number of new aggregationtff.structure
API to expose the Struct
class and relatedtff.profiler
API to expose useful profiling related functions.tff.backends.test
package to expose backends that focused onfederated_secure_sum
intrinsic.tff.experimental
package to expose less stable API.tff.aggregators.AggregationProcessFactory
abstract base classtff.aggregators.UnweightedAggregationFactory
and thetff.aggregators.WeightedAggregationFactory
classes.tff.aggregators.ZeroingFactory
class with atff.aggregators.zeroing_factory
function with the same input arguments.tff.aggregators.ClippingFactory
class with atff.aggregators.clipping_factory
function with the same input arguments.tensorflow
package dependency to 2.4.0
.absl-py
package dependency to 0.10
.grpcio
package dependency to 1.32.0
.jaxlib
package dependency at 0.1.55
.numpy
package dependency to 1.19.2
.tensorflow-addons
package dependency to 0.12.0
.tensorflow-model-optimization
package dependency to 0.5.0
.sequence_reduce
intrinsic handling federated types.Published by ZacharyGarrett almost 4 years ago
tff.aggregators
package with interfaces for stateful aggregationtff.simulations.dataset.gldv2
tff.type_clients
and tff.type_at_server
tff.simulations.compose_dataset_computation_with_iterative_process
APItff.learning.build_federated_averaging_process
use_experimental_simulation_loop
argument to better utilize multi-GPUtff.utils.StatefulFn
, replaced by tff.templates.MeasuredProcess
.tff.learning.assign_weights_to_keras_model
OptimizeDataset
ops from tff.tf_computation
s.research/
directory has been moved toinput_spec
argument for tff.learning.from_keras_model
.2.3.0
.0.4.0
.collections.namedtuple._asdict
calls in collections.OrderedDict
totff.TensorType
with unknown shapes.projector
plugin for tensorboard (https://github.com/tensorflow/federated/issues/914).This release contains contributions from many people at Google, as well as:
@amitport, @ronaldseoh
Published by michaelreneer about 4 years ago
list
s being all_equal
values.Published by michaelreneer about 4 years ago
AnonymousTuple
.AnonymousTuple
to Struct
.Published by michaelreneer over 4 years ago
tensorflow-addons
package dependency to 0.9.0
.tff.backends.native.*
for more information.tff.framework.ExecutionContext
API andtff.templates.MeasuredProcess
concept, a specializationtff.templates.IterativeProcess
.tff.learning
interfaces to accept tff.templates.MeasuredProcess
tff.learning.weights_type_from_model
.tff.framework.FederatingStrategy
, whichtff.framework.FederatingExecutor
so that theIterativeProcess
return fromtff.learning.build_federated_averaging_process
andtff.learning.build_federated_sgd_process
now zip the second tuple outputtff.framework.set_default_executor
function, instead you shouldtff.backends.native.set_local_execution_context
tff.framework.set_default_context
.tff.Computation
base class now contains an abstract __hash__
method,tff.tf_computation
decorated function used no variables.This release contains contributions from many people at Google, as well as:
jvmcns@
Published by jkr26 over 4 years ago
tff.templates.MeasuredProcess
specialization of IterativeProcess
.tff.templates.IterativeProcess
->tff.backends.mapreduce.CanonicalForm
compiler.tff.utils.IterativeProcess
to tff.templates.IterativeProcess
.tff.learning.TrainableModel
, client optimizers are now argumentstff.learning.build_federated_averaging_process
.This release contains contributions from many people at Google, as well as:
jvmcns@
Published by michaelreneer over 4 years ago
Published by michaelreneer over 4 years ago
absl-py
package dependency to 0.9.0
.h5py
package dependency to 2.8.0
.numpy
package dependency to 1.17.5
.tensorflow-privacy
package dependency to 0.2.2
.dummy_batch
parameter of the tff.learning.from_keras_model
Published by ZacharyGarrett over 4 years ago
2.0.0
to 2.1.0
.0.6.0
to 0.7.0
.18.2
to 19.3
.0.1.3
to 0.2.1
.tff.simulation.datasets.cifar100
.tff.learning.build_personalization_eval
for evaluating modeltff.federated_secure_sum
.tff.learning.build_federated_averaing_process()
now takes aclient_optimizer_fn
and a tff.learning.Model
.tff.learning.TrainableModel
is now deprecated.tff.framework.ExecutorFactory
; alltff.framework...executor_factory
calls now return an instance of thisremote_executor_example
binary which demonstrates using theclose()
method to the Executor, allowing subclasses to proactivelytff.federated_zip
on inputs to other federatedtff.framework.create_local_executor
(and similar methods) totff.framework.local_executor_factory
.federated_apply()
, instead use federated_map()
for alltff.learning.ModelWeights
no longer names the tuple fieldstff.sequence_*
on unplaced types now correctly returns a tff.Value
.tff.sequence_*
.. operations are not implemented yet on the newPublished by cramertj almost 5 years ago
federated_map
now works with both tff.SERVER
and tff.CLIENT
federated_zip
received significant performance improvements and now workscollections.OrderedDict
is now required in many places rather thantff.framework.Tuple
is returned fromIterativeProcess.next
.dir()
in ValueImpl
.Published by michaelreneer almost 5 years ago
grpcio
1.24.3
.Published by michaelreneer almost 5 years ago
federated_sample
aggregation that is used to collect a sample oftensorflow
2.0.0
and tensorflow-addons
0.6.0
insteadsetup.py
for releasing TFFtensorflow-privacy
0.2.0
.attr.s
classes type annotations.Execute
method on tff.framework.ExecutorService
to bePublished by ZacharyGarrett about 5 years ago
examples
package split to simulation
and research
.tff.federated_aggregate
supports accumulate
, merge
and report
Published by krzys-ostrowski about 5 years ago
tff.backends.mapreduce
).ConcreteClientData
, random seeds,tf.data.Dataset
string iterator handle.This release contains contributions from many people at Google, as well as:
Gui-U@, Krishna Pillutla, Sergii Khomenko.
Published by krzys-ostrowski over 5 years ago
examples
directory in the pip package.federated_map_all_equal
in the reference executor.This release contains contributions from many people at Google.
Published by jkr26 over 5 years ago
keras
models.tff.utils.update_state
extended to handle more general state
arguments.tff.utils.federated_min
and tff.utils.federated_max
.client_ids
in create_tf_dataset_from_all_clients
by default torequirements.txt
; in particular, grpcio
andportpicker
.tf.data.experimental.NestedStructure
.This release contains contributions from many people at Google, as well as:
Dheeraj R Reddy, @Squadrick.
Published by michaelreneer over 5 years ago
tensorflow
totf-nightly
dependency.attr
module in TFF type system.tff.framework
interface layer.tff.tf_computation
.tf.keras.model.weights
order.Published by ZacharyGarrett over 5 years ago
tff.simulation.TransformingClientData
API and associated inifinite EMNIST dataset (see tensorflow.org/federated/api_docs/python/tff for details)func
to fn
across the repository (rename some parameters and functions)tff.learning.build_federated_evaluation
tff.learning.ModelWeights.assign_weights_to