def create_thread_debugging_execution_context(default_num_clients: int = 0,
clients_per_thread=1):
"""Creates a simple execution context that executes computations locally."""
factory = executor_stacks.thread_debugging_executor_factory(
default_num_clients=default_num_clients,
clients_per_thread=clients_per_thread,
)
def _debug_compiler(comp):
return compiler.transform_to_native_form(comp, transform_math_to_tf=True)
return execution_context.ExecutionContext(
executor_fn=factory, compiler_fn=_debug_compiler)
def create_thread_debugging_execution_context(num_clients=None,
clients_per_thread=1):
"""Creates a simple execution context that executes computations locally."""
factory = executor_stacks.thread_debugging_executor_factory(
num_clients=num_clients,
clients_per_thread=clients_per_thread,
)
def _debug_compiler(comp):
native_form = compiler.transform_to_native_form(comp)
return compiler.transform_mathematical_functions_to_tensorflow(native_form)
return execution_context.ExecutionContext(
executor_fn=factory, compiler_fn=_debug_compiler)
def _create_concurrent_maxthread_tuples():
tuples = []
for concurrency in range(1, 5):
local_ex_string = 'local_executor_{}_clients_per_thread'.format(concurrency)
ex_factory = executor_stacks.local_executor_factory(
clients_per_thread=concurrency)
tuples.append((local_ex_string, ex_factory, concurrency))
sizing_ex_string = 'sizing_executor_{}_client_thread'.format(concurrency)
ex_factory = executor_stacks.sizing_executor_factory(
clients_per_thread=concurrency)
tuples.append((sizing_ex_string, ex_factory, concurrency))
debug_ex_string = 'debug_executor_{}_client_thread'.format(concurrency)
ex_factory = executor_stacks.thread_debugging_executor_factory(
clients_per_thread=concurrency)
tuples.append((debug_ex_string, ex_factory, concurrency))
return tuples
def test_thread_debugging_executor_constructs_exactly_one_reference_resolving_executor(
self, executor_mock):
executor_stacks.thread_debugging_executor_factory().create_executor(
{placements.CLIENTS: 10})
executor_mock.assert_called_once()
class ExecutorStacksTest(parameterized.TestCase):
@parameterized.named_parameters(
('local_executor', executor_stacks.local_executor_factory),
('sizing_executor', executor_stacks.sizing_executor_factory),
('debug_executor', executor_stacks.thread_debugging_executor_factory),
)
def test_construction_with_no_args(self, executor_factory_fn):
executor_factory_impl = executor_factory_fn()
self.assertIsInstance(executor_factory_impl,
executor_stacks.ResourceManagingExecutorFactory)
@parameterized.named_parameters(
('local_executor', executor_stacks.local_executor_factory),
('sizing_executor', executor_stacks.sizing_executor_factory),
)
def test_construction_raises_with_max_fanout_one(self,
executor_factory_fn):
with self.assertRaises(ValueError):
executor_factory_fn(max_fanout=1)
@parameterized.named_parameters(
('local_executor_none_clients',
executor_stacks.local_executor_factory()),
('sizing_executor_none_clients',
executor_stacks.sizing_executor_factory()),
('local_executor_three_clients',
executor_stacks.local_executor_factory(num_clients=3)),
('sizing_executor_three_clients',
executor_stacks.sizing_executor_factory(num_clients=3)),
)
@test_utils.skip_test_for_multi_gpu
def test_execution_of_temperature_sensor_example(self, executor):
comp = _temperature_sensor_example_next_fn()
to_float = lambda x: tf.cast(x, tf.float32)
temperatures = [
tf.data.Dataset.range(10).map(to_float),
tf.data.Dataset.range(20).map(to_float),
tf.data.Dataset.range(30).map(to_float),
]
threshold = 15.0
with executor_test_utils.install_executor(executor):
result = comp(temperatures, threshold)
self.assertAlmostEqual(result, 8.333, places=3)
@parameterized.named_parameters(
('local_executor', executor_stacks.local_executor_factory),
('sizing_executor', executor_stacks.sizing_executor_factory),
)
def test_execution_with_inferred_clients_larger_than_fanout(
self, executor_factory_fn):
@computations.federated_computation(
computation_types.at_clients(tf.int32))
def foo(x):
return intrinsics.federated_sum(x)
executor = executor_factory_fn(max_fanout=3)
with executor_test_utils.install_executor(executor):
result = foo([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
self.assertEqual(result, 55)
@parameterized.named_parameters(
('local_executor_none_clients',
executor_stacks.local_executor_factory()),
('sizing_executor_none_clients',
executor_stacks.sizing_executor_factory()),
('debug_executor_none_clients',
executor_stacks.thread_debugging_executor_factory()),
('local_executor_one_client',
executor_stacks.local_executor_factory(num_clients=1)),
('sizing_executor_one_client',
executor_stacks.sizing_executor_factory(num_clients=1)),
('debug_executor_one_client',
executor_stacks.thread_debugging_executor_factory(num_clients=1)),
)
def test_execution_of_tensorflow(self, executor):
@computations.tf_computation
def comp():
return tf.math.add(5, 5)
with executor_test_utils.install_executor(executor):
result = comp()
self.assertEqual(result, 10)
@parameterized.named_parameters(*_create_concurrent_maxthread_tuples())
def test_limiting_concurrency_constructs_one_eager_executor(
self, ex_factory, clients_per_thread, tf_executor_mock):
num_clients = 10
ex_factory.create_executor({placements.CLIENTS: num_clients})
concurrency_level = math.ceil(num_clients / clients_per_thread)
args_list = tf_executor_mock.call_args_list
# One for server executor, one for unplaced executor, concurrency_level for
# clients.
self.assertLen(args_list, concurrency_level + 2)
@mock.patch(
'tensorflow_federated.python.core.impl.executors.reference_resolving_executor.ReferenceResolvingExecutor',
return_value=ExecutorMock())
def test_thread_debugging_executor_constructs_exactly_one_reference_resolving_executor(
self, executor_mock):
executor_stacks.thread_debugging_executor_factory().create_executor(
{placements.CLIENTS: 10})
executor_mock.assert_called_once()
@parameterized.named_parameters(
('local_executor', executor_stacks.local_executor_factory),
('sizing_executor', executor_stacks.sizing_executor_factory),
('debug_executor', executor_stacks.thread_debugging_executor_factory),
)
def test_create_executor_raises_with_wrong_cardinalities(
self, executor_factory_fn):
executor_factory_impl = executor_factory_fn(num_clients=5)
cardinalities = {
placements.SERVER: 1,
None: 1,
placements.CLIENTS: 1,
}
with self.assertRaises(ValueError, ):
executor_factory_impl.create_executor(cardinalities)
