def test_build_encoded_sum_raise_warning(self):
value = tf.constant(np.random.rand(20))
value_spec = tf.TensorSpec(value.shape, tf.dtypes.as_dtype(value.dtype))
encoder = te.encoders.as_gather_encoder(te.encoders.identity(), value_spec)
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter('always')
encoding_utils.build_encoded_sum(value, encoder)
self.assertLen(w, 1)
def test_build_encoded_sum_raise_warning(self):
value = tf.constant(np.random.rand(20))
value_spec = tf.TensorSpec(value.shape, tf.dtypes.as_dtype(value.dtype))
encoder = te.encoders.as_gather_encoder(te.encoders.identity(), value_spec)
with warnings.catch_warnings(record=True):
warnings.simplefilter('error', DeprecationWarning)
with self.assertRaisesRegex(DeprecationWarning,
'tff.utils.build_encoded_sum()'):
encoding_utils.build_encoded_sum(value, encoder)
def test_run_encoded_sum(self):
value = np.array([0.0, 1.0, 2.0, -1.0])
value_spec = tf.TensorSpec(value.shape,
tf.dtypes.as_dtype(value.dtype))
value_type = computation_types.to_type(value_spec)
encoder = te.encoders.as_gather_encoder(te.encoders.identity(),
value_spec)
gather_fn = encoding_utils.build_encoded_sum(value, encoder)
initial_state = gather_fn.initialize()
@computations.federated_computation(
computation_types.FederatedType(
gather_fn._initialize_fn.type_signature.result,
placements.SERVER),
computation_types.FederatedType(value_type, placements.CLIENTS))
def call_gather(state, value):
return gather_fn(state, value)
_, value_sum = call_gather(initial_state, [value, value])
self.assertAllClose(2 * value, value_sum)
_, value_sum = call_gather(initial_state, [value, -value])
self.assertAllClose(0 * value, value_sum)
_, value_sum = call_gather(initial_state, [value, 2 * value])
self.assertAllClose(3 * value, value_sum)
def build_encoded_sum_from_model(
model_fn: _ModelConstructor,
encoder_fn: _EncoderConstructor) -> computation_utils.StatefulAggregateFn:
"""Builds `StatefulAggregateFn` for weights of model returned by `model_fn`.
This method creates a `GatherEncoder` for every trainable weight of model
created by `model_fn`, as returned by `encoder_fn`.
Args:
model_fn: A Python callable with no arguments function that returns a
`tff.learning.Model`.
encoder_fn: A Python callable with a single argument, which is expected to
be a `tf.Tensor` of shape and dtype to be encoded. The function must
return a `tensor_encoding.core.SimpleEncoder`, which expects a `tf.Tensor`
with compatible type as the input to its `encode` method.
Returns:
A `StatefulAggregateFn` for encoding and summing the weights of model
created by `model_fn`.
Raises:
TypeError: If `model_fn` or `encoder_fn` are not callable objects.
"""
warnings.warn(
'Deprecation warning: '
'tff.learning.framework.build_encoded_sum_from_model() is deprecated, '
'use tff.learning.framework.build_encoded_sum_process_from_model() '
'instead.', DeprecationWarning)
py_typecheck.check_callable(model_fn)
py_typecheck.check_callable(encoder_fn)
trainable_weights = _weights_from_model_fn(model_fn).trainable
encoders = tf.nest.map_structure(encoder_fn, trainable_weights)
return encoding_utils.build_encoded_sum(trainable_weights, encoders)
def test_build_encoded_sum(self, value_constructor, encoder_constructor):
value = value_constructor(np.random.rand(20))
value_spec = tf.TensorSpec(value.shape,
tf.dtypes.as_dtype(value.dtype))
value_type = computation_types.to_type(value_spec)
encoder = te.encoders.as_gather_encoder(encoder_constructor(),
value_spec)
gather_fn = encoding_utils.build_encoded_sum(value, encoder)
self.assertIsInstance(gather_fn, StatefulAggregateFn)
state_type = gather_fn._initialize_fn.type_signature.result
@computations.federated_computation(
computation_types.FederatedType(state_type, placements.SERVER),
computation_types.FederatedType(value_type, placements.CLIENTS),
computation_types.FederatedType(
computation_types.to_type(tf.float32), placements.CLIENTS))
def gather_computation(state, value, weight):
return gather_fn._next_fn(state, value, weight)
gather_result_type = gather_computation.type_signature.result
self.assertEqual(state_type, gather_result_type[0].member)
self.assertEqual(placements.SERVER, gather_result_type[0].placement)
self.assertEqual(value_type, gather_result_type[1].member)
self.assertEqual(placements.SERVER, gather_result_type[1].placement)
def test_build_encoded_sum_raises_bad_structure(self):
value = [tf.constant([0.0, 1.0]), tf.constant([0.0, 1.0])]
encoder = te.encoders.as_gather_encoder(te.encoders.identity(),
tf.TensorSpec((2,)))
with self.assertRaises(ValueError):
encoding_utils.build_encoded_sum(value, encoder)
def test_build_encoded_sum_raises_incompatible_encoder(self):
value = tf.constant([0.0, 1.0])
incompatible_encoder = te.encoders.as_gather_encoder(
te.encoders.identity(), tf.TensorSpec((3,)))
with self.assertRaises(TypeError):
encoding_utils.build_encoded_sum(value, incompatible_encoder)
def test_build_encoded_sum_raises_bad_encoder(self, bad_encoder):
value = tf.constant([0.0, 1.0])
with self.assertRaises(TypeError):
encoding_utils.build_encoded_sum(value, bad_encoder)
