Skip to content
Snippets Groups Projects
Select Git revision
  • cl/332000035
  • cl/343353741
  • cl/346635895
  • cl/347081719
  • cl/347401844
  • cl/347907220
  • cl/348027488
  • cl/348050964
  • cl/348070543
  • cl/348098438
  • cl/348138051
  • cl/348556585
  • cl/348851214
  • cl/348866975
  • cl/349624352
  • cl/350701164
  • cl/352033008
  • cl/353341137
  • cl/354238630
  • cl/354591074
  • v0.68.0
  • v0.67.0
  • v0.66.0
  • v0.65.0
  • v0.64.0
  • v0.63.0
  • v0.62.0
  • v0.61.0
  • v0.60.0
  • v0.59.0
  • v0.58.0
  • v0.57.0
  • v0.56.0
  • v0.55.0
  • v0.54.0
  • v0.53.0
  • v0.52.0
  • v0.51.0
  • v0.50.0
  • v0.49.0
40 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.015Mar1110985432128Feb2726252423222119181718171613121110965432129Jan2827262522211915141312118765430Dec292824232221171615141198753130Nov2726252423201918171612111096543230Oct292827232221201917161514131211987652129Sep2825242322211916151413121110986432128Aug272625212019181715141312107Remove `impl` package visibility exception.Removes rpc_mode argument from TFF remove executor examples and tutorials.Restrict visibility rules in `tools` package.Update gcp_setup.mdAlign the checkpoint and metrics manager API.Uses LRU cache for executor factory internal caches.Replaces remove_called_lambdas_and_blocks with transform_to_local_call_dominant.Add type_to_tensor_structureCopy ClientData class and subclasses to tff.simulation.datasets.Add broadcast_process argument to build_federated_evaluation to allow passing in a MeasuredProcess that can encode/decode during eval tasks similar to what is done for train tasks. The broadcast_process must not require state; for example, a uniform_quantization encoder can be used to compress the model that is being sent from server to clients for evaluation.InternalAdd optional rounds_per_checkpoint argument to enable periodic checkpoint saving without the need to hardcode round number checks into a training loop.Create a LearningProcess template specializing IterativeProcess.Add `.bazelversion` configuration.Adds possibility of unweighted aggregation to tff.learning.secure_aggregator.Adds possibility of unweighted aggregation to tff.learning.compression_aggregator.Skips the first notebook cell when testing.Adds possibility of unweighted aggregation to tff.learning.robust_aggregator.Expose `UnweightedReservoirSamplingFactory` in the `tff.aggregators` package.Expose `SqlClientData` in the pip package public API.Remove selection-by-name support from executorsUse zip for grads and vars in apply_gradients in simple_fedavg.Adds more detailed description of the effects of the `clipping` and `zeroing` arguments to aggregators in model_update_aggregator.py.Add an UnweightedAggregationFactory that performs unweighted reservoir sampling using `tff.federated_aggregate`.Fix a typo in `update_state()` description.Update tff docs to describe tff.aggregators for differentially private aggregation.Finish cleaning up the `impl` package.Move the `tree_to_cc_transformations_test` module to the `impl` package.Extend `tff.utils.update_state` for work on `tff.structure.Struct` inputs.Updates TFF callsites for graphdef equality.Flatten selections to index in computation.protoTransformingClientData does not add suffixes if not creating pseudo-clients.Adds a small tutorial on JAX support.Remove unused GCP endpoint scripts.Automated rollback of commit bf7f333585591101ce578b782d3cf7dbf45db2a5TransformingClientData defaults to having same number of clients as original data.Adds AggregationProcess.is_weighted property.Uses new AggregationProcess.is_weighted property.cl/359576676 up…cl/359576676 upstream/cl/359576676Factors out federated averaging for JAX as a reusable component to significantly shorten the JAX/XLA training example.Adds download and other links as appropriate to TFF tutorials.
Loading