An open-source framework for machine learning and other computations on decentralized data.
APACHE-2.0 License
Bot releases are hidden (Show)
Published by zcharles8 almost 2 years ago
tff.learning.templates.build_apply_optimizer_finalizer
.tff.learning.framework
functools.wraps
within tff.tf_computation
.Published by michaelreneer almost 2 years ago
tff.learning.models.FunctionModel
support to all methods intff.learning.algorithms
.tf.data.DataSpec
to tff.types.infer_unplaced_type
.tensorflow::ThreadPool
for parallelism inside the C++TensorFlowExecutor
.tff.experimental_tf_fn_computation
tracing decorator thatFunctionDef
instead of GraphDef
tracing, providing tf.function
number_of_clients
to num_clients
in the federated program API.tff.learning.templates
.
tff.learning.framework.build_model_delta_optimizer_process
tff.learning.framework.ClientDeltaFn
Published by ZacharyGarrett about 2 years ago
tff.learning.models.FunctionalModel
support totff.learning.algorithms.build_mime_lite
.tensorflow-privacy
to version 0.8.6
.tff.framework.Context
.tff.simulation.datasets.gldv2.get_synthetic
.tff.program.PrefetchingDataSource
.tff.learning.framework.build_encoded_broadcast_process_from_model
.tff.learning.ModelWeights
and aliastff.learning.framework.ModelWeights
, has now moved totff.learning.models.ModelWeights
. Code should be updated before the nexttff.learning.models.functional_model_from_keras
.tff.tf_computation
tracing to also trace functools.partial
Published by wushanshan about 2 years ago
numpy
values in the tff.program
API.tff.learning.ModelWeights.convert_variables_to_arrays
.tff.learning.metrics
.tff.framework.ComposingExecutorFactory
.jax
and jaxlib
version 0.3.14
.tff.program.CoroValueReference
totff.program.AwaitableValueReference
to reflect the relaxed contract.tff.simulation.build_uniform_sampling_fn
,tff.learning.robust_aggregator
,tff.aggregators.PrivateQuantileEstimationProcess
.tff.simulation.baselines.BaselineTask
have the same element structure.This release contains contributions from many people at Google, as well as:
Madhava Jay, nbishdev@
Published by michaelreneer about 2 years ago
tff.learning.models.FunctionalModel
totff.learning.algorithms.build_fed_sgd
andtff.learning.algorithms.build_fed_prox
.tensorflow
to version 2.10
.tff.analytics.heavy_hitters.iblt.build_iblt_computation()
.tff.federated_select
.Published by michaelreneer about 2 years ago
tff.learning.templates.ClientWorkProcess
.tff.learning.algorithms.build_mime_lite_with_optimizer_schedule
.tensorflow-privacy
to version 0.8.5
.tff.learning.entropy_compression_aggregator
.tff.aggregators.EliasGammaEncodedSumFactory
.tff.program.ClientIdDataSource
andtff.program.ClientIdDataSourceIterator
, for working with a data source ofloading_remote_data
tutorial.tff.simulation.datasets.stackoverflow.get_synthetic
.Published by jkr26 about 2 years ago
Bazel
version 5.3.0
.setup.py
to explicitly fail to pip install
in Python 3.10.pip
.FederatedDataSource
.type_signature
attribute to objects of type MapReduceForm
.0.8.4
.0.8.3
.initialize
from MapReduceForm
.SequenceType
now automatically casts any StructWithPythonType
thatlist
to a tuple
for tf.data
compatibility.model_fn
and model
parameters oftff.learning.algorithms.build_weighted_fed_avg
.MapReduceForm
now takes a type_signature
argument in its constructor,initialize
argument.MapReduceForm
no longer contains an initialize
attribute.Published by ZacharyGarrett about 2 years ago
tff.analytics.heavy_hitters.iblt
with create_chunker
API fortff.aggregators.DifferentiallyPrivateFactory.tree_aggregation
withrecord_aggregation_factory
argument.ModularClippingSumFactory
with SecureModularSumFactory
intff.analytics.build_hierarchical_histogram_process
.Published by ZacharyGarrett about 2 years ago
tff.learning.algorithms.build_mime_lite_with_optimizer_schedule
.Published by hardik-vala about 2 years ago
ReleaseManager
s to make authoring program logic more convenient.attrs
dependency to version 21.4.0
.tensorflow-privacy
dependency to version 0.8.1
.tff.learning.BatchOutput
from an attrs class to a namedtuple.tff.learning.framework.parameter_count_from_model
API.Published by wushanshan over 2 years ago
namedtuple
s in the tff.program
package.num_subrounds
parameter to the mergeable context, allowing callers totff.learning.models.FunctionalModel
, includingcreate_functional_metric_fns
and thetff.learning.algorithms.build_weighted_fed_avg
.tff.program
package.tff.structure.Struct
integration with JAX pytrees to not flattenPublished by xiaoyux11 over 2 years ago
MemoryReleaseManager
to save type_signature
when releasingtype_signature
parameter to the ReleaseManager.release
method.python_executor_stacks.py
inexecutor_stacks
directory.metric_finalizers
use metric constructors if available.Published by jkr26 over 2 years ago
tff.learning.algorithms
API.tff.learning.build_federated_averaging_process
; users shouldtff.learning.algorithms.build_weighted_fed_avg
.Dispose
, DisposeExecutor
,max_concurrent_computation_calls
.MergeableCompForm
, removing spurious failures on dependent-aggregationtff.framework.deserialize_value
may fail to deserializetff.framework.serialize_value
may fail to serialize tf.data.DatasetsPublished by hardik-vala over 2 years ago
DataBackend
to load remote datasets.Published by hbmcmahan over 2 years ago
2.9.1
.2.9.2
.MeasuredProcess
es.program
package._GLIBCXX_USE_CXX11_ABI
to 1
.Published by wennanzhu over 2 years ago
DataBackend
with TFF comps.tff.test.TestCase
, tff.test.main()
, and deletetest_case
module.compiler/test_utils
to compiler/building_block_test_utils
.pytype: skip-file
.None
return type of ReleaseManager.release
.random_integers
with randint
.Published by wushanshan over 2 years ago
asyncio.run
call to metrics manager release calls to ensuremodel_update_aggregator
to support structures with mixed floating dtypes.tff.simulation.compose_dataset_computation_with_iterative_process
for tff.learning.templates.LearningProcess
.tff.test.TestCase
to tff.test.*
as functions.assert_type_assignable_from
function.assert_nested_struct_eq
to the type_conversions_test
module.client_train_process
and fedavg_ds_loop comparisons.async_utils.SharedAwaitable
exception raiser.Published by xiaoyux11 over 2 years ago
tff.learning.build_federated_averaging_process
.tf.keras.metrics.Metric
to a set of puretf.functions
.ProgramStateManager.version
to ProgramStateManager.get_versions
.Published by michaelreneer over 2 years ago
5.1.1
.tff.program
API to use asyncio
.tff.framework
package:
tff.framework.CardinalitiesType
.tff.framework.PlacementLiteral
.tff.framework.merge_cardinalities
.tff.analytics
analytic_gauss_stddev
API.ProgramStateManager.version
to ProgramStateManager.get_versions
.Published by ZacharyGarrett over 2 years ago
tff.analytics
tff.analytics.IbltFactory
aggregation factory.tff.analytics.heavy_hitters.iblt.build_iblt_computation
.tff.aggregator
as_weighted_aggregator
to the tff.aggregator.Factory
API.tff.learning
tff.learning.metrics.secure_sum_then_finalize
by grouping tensors byset_model_weights
method and default implementation totff.learning.templates.LearningProcess
.reset_metrics
attribute to tff.learning.Model
.schedule_learning_rate
to tff.learning.optimizers
.tff.learning.ddp_secure_aggregator
for Distributedtff.simulation
report_local_outputs
andfederated_output_computation
from tff.learning.Model
ingest
method from tff.Context
CLIENTS
cardinality is zero doesn't work on the Pythonmodel_fn