syft.frameworks.keras.model.sequential

Module Contents

syft.frameworks.keras.model.sequential._args_not_supported_by_tfe = ['activity_regularizer', 'kernel_regularizer', 'bias_regularizer', 'kernel_constraint', 'bias_constraint', 'dilation_rate']

syft.frameworks.keras.model.sequential.share(model, cluster, target_graph=None)

Secret share the model between workers.

This is done by rebuilding the model as a TF Encrypted model inside target_graph and pushing this graph to TFEWorkers running the cluster.

Note that this keeps a TensorFlow session alive that must later be shutdown using model.stop().

syft.frameworks.keras.model.sequential.serve(model, num_requests=5)

Serve the specified number of predictions using the shared model, blocking until completed.

syft.frameworks.keras.model.sequential.stop(model)

Shutdown the TensorFlow session that was used to serve the shared model.

syft.frameworks.keras.model.sequential._configure_tfe(cluster)

syft.frameworks.keras.model.sequential._rebuild_tfe_model(keras_model, stored_keras_weights)

Rebuild the plaintext Keras model as a TF Encrypted Keras model from the plaintext weights in stored_keras_weights using the current TensorFlow graph, and the current TF Encrypted protocol and configuration.

syft.frameworks.keras.model.sequential._instantiate_tfe_layer(keras_layer, stored_keras_weights)

syft.frameworks.keras.model.sequential._get_layer_type(keras_layer_cls)

syft.frameworks.keras.model.sequential._trim_params(params, filter_list)