def server_update_fn(server_state, sub_round_votes):
return server_update(server_state,
tf.constant(possible_prefix_extensions),
sub_round_votes,
num_sub_rounds=tf.constant(num_sub_rounds),
max_num_prefixes=tf.constant(max_num_prefixes),
threshold=tf.constant(threshold))
def test_server_update_finds_heavy_hitters_with_threshold(self):
max_num_prefixes = tf.constant(10)
threshold = tf.constant(5)
num_sub_rounds = tf.constant(1, dtype=tf.int32)
possible_prefix_extensions = [
'a', 'b', 'c', 'd', triehh_tf.DEFAULT_TERMINATOR
]
possible_prefix_extensions = tf.constant(possible_prefix_extensions,
dtype=tf.string)
discovered_prefixes = [
'a', 'b', 'c', 'd', triehh_tf.DEFAULT_TERMINATOR
]
discovered_heavy_hitters = []
heavy_hitters_counts = []
server_state = triehh_tf.ServerState(
discovered_heavy_hitters=tf.constant(discovered_heavy_hitters,
dtype=tf.string),
heavy_hitters_counts=tf.constant(heavy_hitters_counts,
dtype=tf.int32),
discovered_prefixes=tf.constant(discovered_prefixes,
dtype=tf.string),
round_num=tf.constant(1, dtype=tf.int32),
accumulated_votes=tf.zeros(
dtype=tf.int32,
shape=[max_num_prefixes,
len(possible_prefix_extensions)]))
sub_round_votes = tf.constant(
[[10, 9, 8, 7, 6], [5, 4, 3, 0, 4], [2, 1, 0, 0, 0], [
0, 0, 0, 0, 0
], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0],
[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0]],
dtype=tf.int32)
server_state = triehh_tf.server_update(server_state,
possible_prefix_extensions,
sub_round_votes, num_sub_rounds,
max_num_prefixes, threshold)
expected_discovered_prefixes = tf.constant(
['aa', 'ab', 'ac', 'ad', 'ba'], dtype=tf.string)
expected_discovered_heavy_hitters = tf.constant(['a'], dtype=tf.string)
expected_heavy_hitters_counts = tf.constant([6], dtype=tf.int32)
expected_accumulated_votes = tf.constant(
[[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [
0, 0, 0, 0, 0
], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0],
[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0]],
dtype=tf.int32)
self.assertSetAllEqual(server_state.discovered_prefixes,
expected_discovered_prefixes)
self.assertSetAllEqual(server_state.discovered_heavy_hitters,
expected_discovered_heavy_hitters)
self.assertHistogramsEqual(server_state.discovered_heavy_hitters,
server_state.heavy_hitters_counts,
expected_discovered_heavy_hitters,
expected_heavy_hitters_counts)
self.assertAllEqual(server_state.accumulated_votes,
expected_accumulated_votes)
def server_update_fn(server_state, sub_round_votes):
server_state = server_update(
server_state,
sub_round_votes,
num_sub_rounds=tf.constant(num_sub_rounds),
max_num_heavy_hitters=tf.constant(max_num_heavy_hitters),
default_terminator=tf.constant(default_terminator, dtype=tf.string))
return server_state
def test_server_update_does_not_decode_in_a_subround(self):
max_num_prefixes = tf.constant(10)
threshold = tf.constant(1)
num_sub_rounds = tf.constant(2, dtype=tf.int32)
possible_prefix_extensions = [
'a', 'b', 'c', 'd', 'e', triehh_tf.DEFAULT_TERMINATOR
]
possible_prefix_extensions = tf.constant(
possible_prefix_extensions, dtype=tf.string)
discovered_prefixes = ['']
discovered_heavy_hitters = []
heavy_hitters_frequencies = []
server_state = triehh_tf.ServerState(
discovered_heavy_hitters=tf.constant(
discovered_heavy_hitters, dtype=tf.string),
heavy_hitters_frequencies=tf.constant(
heavy_hitters_frequencies, dtype=tf.float64),
discovered_prefixes=tf.constant(discovered_prefixes, dtype=tf.string),
round_num=tf.constant(0, dtype=tf.int32),
accumulated_votes=tf.zeros(
dtype=tf.int32,
shape=[max_num_prefixes,
len(possible_prefix_extensions)]),
accumulated_weights=tf.constant(0, dtype=tf.int32))
sub_round_votes = tf.constant(
[[1, 2, 1, 2, 0, 0], [2, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0]],
dtype=tf.int32)
sub_round_weight = tf.constant(10, dtype=tf.int32)
server_state = triehh_tf.server_update(server_state,
possible_prefix_extensions,
sub_round_votes, sub_round_weight,
num_sub_rounds, max_num_prefixes,
threshold)
expected_discovered_prefixes = tf.constant([''], dtype=tf.string)
expected_discovered_heavy_hitters = tf.constant([], dtype=tf.string)
expected_heavy_hitters_frequencies = tf.constant([], dtype=tf.float64)
expected_accumulated_votes = sub_round_votes
expected_accumulated_weights = sub_round_weight
self.assertSetAllEqual(server_state.discovered_prefixes,
expected_discovered_prefixes)
self.assertSetAllEqual(server_state.discovered_heavy_hitters,
expected_discovered_heavy_hitters)
self.assertHistogramsEqual(server_state.discovered_heavy_hitters,
server_state.heavy_hitters_frequencies,
expected_discovered_heavy_hitters,
expected_heavy_hitters_frequencies)
self.assertAllEqual(server_state.accumulated_votes,
expected_accumulated_votes)
self.assertEqual(server_state.accumulated_weights,
expected_accumulated_weights)
def test_server_update_finds_heavy_hitters(self):
max_num_heavy_hitters = tf.constant(10)
default_terminator = tf.constant('$', tf.string)
num_sub_rounds = tf.constant(1, dtype=tf.int32)
possible_prefix_extensions = ['a', 'b', 'c', 'd', '$']
discovered_prefixes = ['a', 'b', 'c', 'd', '$']
discovered_heavy_hitters = []
server_state = triehh_tf.ServerState(
discovered_heavy_hitters=tf.constant(discovered_heavy_hitters,
dtype=tf.string),
discovered_prefixes=tf.constant(discovered_prefixes,
dtype=tf.string),
possible_prefix_extensions=tf.constant(possible_prefix_extensions,
dtype=tf.string),
round_num=tf.constant(1, dtype=tf.int32),
accumulated_votes=tf.zeros(
dtype=tf.int32,
shape=[max_num_heavy_hitters,
len(possible_prefix_extensions)]))
sub_round_votes = tf.constant(
[[10, 9, 8, 7, 6], [5, 4, 3, 0, 0], [2, 1, 0, 0, 0], [
0, 0, 0, 0, 0
], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0],
[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0]],
dtype=tf.int32)
server_state = triehh_tf.server_update(server_state, sub_round_votes,
num_sub_rounds,
max_num_heavy_hitters,
default_terminator)
expected_discovered_prefixes = tf.constant(
['aa', 'ab', 'ac', 'ad', 'ba', 'bb', 'bc', 'ca', 'cb'],
dtype=tf.string)
expected_discovered_heavy_hitters = tf.constant(['a'], dtype=tf.string)
expected_accumulated_votes = tf.constant(
[[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [
0, 0, 0, 0, 0
], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0],
[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0]],
dtype=tf.int32)
self.assertAllEqual(server_state.discovered_prefixes,
expected_discovered_prefixes)
self.assertAllEqual(server_state.discovered_heavy_hitters,
expected_discovered_heavy_hitters)
self.assertAllEqual(server_state.accumulated_votes,
expected_accumulated_votes)
def test_server_update_works_on_empty_discovered_prefixes(self):
max_num_prefixes = tf.constant(10)
threshold = tf.constant(1)
num_sub_rounds = tf.constant(1, dtype=tf.int32)
possible_prefix_extensions = [
'a', 'b', 'c', 'd', 'e', triehh_tf.DEFAULT_TERMINATOR
]
possible_prefix_extensions = tf.constant(possible_prefix_extensions,
dtype=tf.string)
discovered_prefixes = []
discovered_heavy_hitters = []
server_state = triehh_tf.ServerState(
discovered_heavy_hitters=tf.constant(discovered_heavy_hitters,
dtype=tf.string),
discovered_prefixes=tf.constant(discovered_prefixes,
dtype=tf.string),
round_num=tf.constant(1, dtype=tf.int32),
accumulated_votes=tf.zeros(
dtype=tf.int32,
shape=[max_num_prefixes,
len(possible_prefix_extensions)]))
sub_round_votes = tf.constant(
[[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0]],
dtype=tf.int32)
server_state = triehh_tf.server_update(server_state,
possible_prefix_extensions,
sub_round_votes, num_sub_rounds,
max_num_prefixes, threshold)
expected_discovered_prefixes = tf.constant([], dtype=tf.string)
expected_discovered_heavy_hitters = tf.constant([], dtype=tf.string)
expected_accumulated_votes = tf.constant(
[[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0]],
dtype=tf.int32)
self.assertSetAllEqual(server_state.discovered_prefixes,
expected_discovered_prefixes)
self.assertSetAllEqual(server_state.discovered_heavy_hitters,
expected_discovered_heavy_hitters)
self.assertAllEqual(server_state.accumulated_votes,
expected_accumulated_votes)
def test_server_update_does_not_decode_in_a_subround(self):
max_num_heavy_hitters = tf.constant(10)
default_terminator = tf.constant('$', tf.string)
num_sub_rounds = tf.constant(2, dtype=tf.int32)
possible_prefix_extensions = ['a', 'b', 'c', 'd', 'e']
discovered_prefixes = ['']
discovered_heavy_hitters = []
server_state = triehh_tf.ServerState(
discovered_heavy_hitters=tf.constant(discovered_heavy_hitters,
dtype=tf.string),
discovered_prefixes=tf.constant(discovered_prefixes,
dtype=tf.string),
possible_prefix_extensions=tf.constant(possible_prefix_extensions,
dtype=tf.string),
round_num=tf.constant(0, dtype=tf.int32),
accumulated_votes=tf.zeros(
dtype=tf.int32,
shape=[max_num_heavy_hitters,
len(possible_prefix_extensions)]))
sub_round_votes = tf.constant(
[[1, 2, 1, 2, 0], [2, 0, 0, 0, 0], [0, 0, 0, 0, 0], [
0, 0, 0, 0, 0
], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0],
[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [0, 0, 0, 0, 0]],
dtype=tf.int32)
server_state = triehh_tf.server_update(server_state, sub_round_votes,
num_sub_rounds,
max_num_heavy_hitters,
default_terminator)
expected_discovered_prefixes = tf.constant([''], dtype=tf.string)
expected_discovered_heavy_hitters = tf.constant([], dtype=tf.string)
expected_accumulated_votes = sub_round_votes
self.assertAllEqual(server_state.discovered_prefixes,
expected_discovered_prefixes)
self.assertAllEqual(server_state.discovered_heavy_hitters,
expected_discovered_heavy_hitters)
self.assertAllEqual(server_state.accumulated_votes,
expected_accumulated_votes)
def test_all_tf_functions_work_together(self):
clients = 3
num_sub_rounds = 4
max_rounds = 6
max_num_heavy_hitters = 3
max_user_contribution = 100
roots = (string.ascii_lowercase + string.digits + "'@#-;*:./" +
triehh_tf.DEFAULT_TERMINATOR)
possible_prefix_extensions = list(roots)
server_state = triehh_tf.ServerState(
discovered_heavy_hitters=tf.constant([], dtype=tf.string),
discovered_prefixes=tf.constant([''], dtype=tf.string),
possible_prefix_extensions=tf.constant(possible_prefix_extensions,
dtype=tf.string),
round_num=tf.constant(0, dtype=tf.int32),
accumulated_votes=tf.zeros(
dtype=tf.int32,
shape=[max_num_heavy_hitters,
len(possible_prefix_extensions)]))
def create_dataset_fn(client_id):
del client_id
return tf.data.Dataset.from_tensor_slices(['hello', 'hey', 'hi'])
client_ids = list(range(100))
client_data = tff.simulation.ClientData.from_clients_and_fn(
client_ids=client_ids,
create_tf_dataset_for_client_fn=create_dataset_fn)
for round_num in range(max_rounds * num_sub_rounds):
sampled_clients = list(range(clients))
sampled_datasets = [
client_data.create_tf_dataset_for_client(client_id)
for client_id in sampled_clients
]
accumulated_votes = tf.zeros(
dtype=tf.int32,
shape=[max_num_heavy_hitters,
len(possible_prefix_extensions)])
# This is a workaround to clear the graph cache in the `tf.function`; this
# is necessary because we need to construct a new lookup table every round
# based on new prefixes.
client_update = tf.function(
triehh_tf.client_update.python_function)
for dataset in sampled_datasets:
client_output = client_update(
dataset, server_state.discovered_prefixes,
server_state.possible_prefix_extensions, round_num,
tf.constant(num_sub_rounds),
tf.constant(max_num_heavy_hitters, dtype=tf.int32),
tf.constant(max_user_contribution, dtype=tf.int32))
accumulated_votes += client_output.client_votes
server_state = triehh_tf.server_update(
server_state, accumulated_votes,
tf.constant(num_sub_rounds, dtype=tf.int32),
tf.constant(max_num_heavy_hitters, dtype=tf.int32),
tf.constant(triehh_tf.DEFAULT_TERMINATOR, dtype=tf.string))
expected_discovered_heavy_hitters = tf.constant(['hi', 'hey', 'hello'],
dtype=tf.string)
self.assertAllEqual(server_state.discovered_heavy_hitters,
expected_discovered_heavy_hitters)