Attend the Women in ML Symposium on December 7 Register now


Stay organized with collections Save and categorize content based on your preferences.

Aggregation Factory that performs secure summation over metrics.

Inherits From: UnweightedAggregationFactory

The created tff.templates.AggregationProcess uses the inner summation processes created by the tff.aggregators.SecureSumFactory to sum unfinalized metrics from tff.CLIENTS to tff.SERVER.

Internally metrics are grouped by their value range and dtype, and only one secure aggregation process will be created for each group. This is an optimization for computation tracing and compiling, which can be slow when there are a large number of independent aggregations.

The initialize function initializes the state for each inner secure aggregation progress. The next function takes the state and local unfinalized metrics reported from tff.CLIENTS, and returns a tff.templates.MeasuredProcessOutput object with the following properties:

  • state: an collections.OrderedDict of the states of the inner secure aggregation processes.
  • result: an collections.OrderedDict of secure summed unfinalized metrics.
  • measurements: an collections.OrderedDict of the measurements of inner secure aggregation processes.

metric_value_ranges An optional collections.OrderedDict that matches the structure of local_unfinalized_metrics_type (a value for each tff.types.TensorType in the type tree). Each leaf in the tree should have a 2-tuple that defines the range of expected values for that variable in the metric. If the entire structure is None, a default range of [0.0, 2.0**20 - 1] will be applied to integer variables and auto-tuned bounds will be applied to float variable. Each leaf may also be None, which will also get the default range according to the variable value type; allowing partial user sepcialization. At runtime, values that fall outside the ranges specified at the leaves will be clipped to within the range.

TypeError If metric_value_ranges type mismatches.



View source

Creates an AggregationProcess for secure summation over metrics.

local_unfinalized_metrics_type A tff.types.StructWithPythonType (with collections.OrderedDict as the Python container) of a client's local unfinalized metrics. Let local_unfinalized_metrics be the output of tff.learning.Model.report_local_unfinalized_metrics(), its type can be obtained by tff.framework.type_from_tensors(local_unfinalized_metrics).

An instance of tff.templates.AggregationProcess.

TypeError If any argument type mismatches.