Structure for outputs returned from clients during federated optimization.
tff.learning.framework.ClientOutput(
weights_delta, weights_delta_weight, model_output,
optimizer_output=attr_dict['optimizer_output'].default
)
Attributes |
|---|
weights_delta
|
A dictionary of updates to the model's trainable
variables.
|
weights_delta_weight
|
Weight to use in a weighted mean when aggregating
weights_delta.
|
model_output
|
A structure matching
tff.learning.Model.report_local_outputs, reflecting the results of
training on the input dataset.
|
optimizer_output
|
Additional metrics or other outputs defined by the
optimizer.
|
View source on GitHub