def _fake_mirrored(value, devices):
"""Create a faked Mirrored object for testing.
All components of the returned Mirrored have the same objects, which is not
true in reality.
"""
devices = _get_devices(devices)
values = []
for d in devices:
with ops.device(d):
values.append(array_ops.identity(value))
return distribute_utils.regroup(values, wrap_class=value_lib.Mirrored)
def testNamedTuple(self):
# We include toy implementations of Scaffold and EstimatorSpec to
# avoid a dependency on Estimator here.
class Scaffold(object):
pass
class EstimatorSpec(
collections.namedtuple(
"EstimatorSpec",
["mode", "loss", "train_op", "scaffold"])):
def __new__(cls, mode, loss, train_op, scaffold=None):
return super(EstimatorSpec, cls).__new__(cls,
mode=mode,
loss=loss,
train_op=train_op,
scaffold=scaffold
or Scaffold())
with context.graph_mode(), ops.Graph().as_default():
created_estimator_specs = []
for device_id in range(3):
spec = EstimatorSpec(mode=mode_keys.EstimatorModeKeys.TRAIN,
loss=constant_op.constant(device_id / 2),
train_op=array_ops.identity(
constant_op.constant(device_id)))
created_estimator_specs.append(spec)
merged_estimator_spec = distribute_utils.regroup(
created_estimator_specs)
self.assertIsInstance(merged_estimator_spec, EstimatorSpec)
self.assertEqual(mode_keys.EstimatorModeKeys.TRAIN,
merged_estimator_spec.mode)
for device_id in range(3):
self.assertEqual(created_estimator_specs[device_id].loss,
merged_estimator_spec.loss.values[device_id])
self.assertEqual(
created_estimator_specs[device_id].train_op,
merged_estimator_spec.train_op.values[device_id])
# Scaffold is populated by `EstimatorSpec.__new__`.
self.assertEqual(
created_estimator_specs[device_id].scaffold,
merged_estimator_spec.scaffold.values[device_id])
self.assertIsInstance(
created_estimator_specs[device_id].scaffold, Scaffold)
# Also test that we can undo the merge using select_replica()
self.assertEqual(
created_estimator_specs[device_id],
distribute_utils.select_replica(device_id,
merged_estimator_spec))
def simple_broadcast(value, destinations, always_mirrored=False):
"""Broadcast `value` to `destinations` using simple copies."""
devices = get_devices_from(destinations)
if len(devices) == 1 and not always_mirrored:
return cross_device_utils.copy_tensor_or_indexed_slices_to_device(
value, devices[0])
else:
value_updates = []
for d in devices:
value_updates.append(
cross_device_utils.copy_tensor_or_indexed_slices_to_device(value, d))
return distribute_utils.regroup(value_updates,
wrap_class=value_lib.Mirrored)
def batch_reduce(self,
reduce_op,
value_destination_pairs,
experimental_hints=None):
"""Reduce PerReplica objects in a batch.
Reduce each first element in `value_destination_pairs` to each second
element which indicates the destinations.
This can be faster than multiple individual `reduce`s because we can
fuse several tensors into one or multiple packs before reduction.
Args:
reduce_op: An instance of `tf.distribute.ReduceOp` that indicates how the
`per_replica_value` will be reduced.
value_destination_pairs: A list or a tuple of PerReplica objects (or
tensors with device set if there is one device) and destinations.
experimental_hints: A `tf.distrbute.experimental.CollectiveHints`. Hints
to perform collective operations.
Returns:
a list of Mirrored objects.
Raises:
ValueError: if `value_destination_pairs` is not an iterable of
tuples of PerReplica objects and destinations.
"""
# TODO(yuefengz): if destinations are different, split into several
# `_batch_reduce` invocations.
if not _validate_value_destination_pairs(value_destination_pairs):
# If the first element of each pair is a tensor, we try to turn it into a
# PerReplica object.
value_destination_pairs = _normalize_value_destination_pairs(
value_destination_pairs)
for _, d in value_destination_pairs:
validate_destinations(d)
# Shortcut all PerReplica objects only contain one value.
if self._num_between_graph_workers == 1 and _all_devices_match(
value_destination_pairs) and len(
value_destination_pairs[0][0].values) == 1:
return [
distribute_utils.regroup(v.values, wrap_class=value_lib.Mirrored)
for v, _ in value_destination_pairs
]
if experimental_hints is None:
experimental_hints = collective_util.Hints()
return self.batch_reduce_implementation(reduce_op, value_destination_pairs,
experimental_hints)
def _call_for_each_replica(self, fn, args, kwargs):
# For now, `fn` must be an @tf.function.
# TODO(josh11b): Relax this restriction? Main problem is if
# (a) executing eagerly, (b) `fn` not @tf.function, and
# (c) executed frequently.
assert isinstance(fn, def_function.Function)
if _outside_run_graph() is not None:
# Nested case, should just use outer function's context for things like
# the current replica index.
# TODO(josh11b): Test this case!
with MirroredFunctionReplicaContext(self._container_strategy()):
results = fn(*nest.map_structure(_unwrap_tensors, args),
**nest.map_structure(_unwrap_tensors, kwargs))
return nest.map_structure(_wrap_tensors, results)
_replica_index.graph_outside_run = ops.get_default_graph()
return_values = []
try:
with MirroredFunctionReplicaContext(self._container_strategy()):
for index, device in enumerate(self._devices):
_replica_index.current = index
with ops.device(device):
if context.executing_eagerly():
# NOTE: These functions need to execute concurrently if they
# use a collective op. This is a particular concern with eager
# execution.
with context.execution_mode(context.ASYNC):
return_values.append(
fn(
*distribute_utils.select_replica(
index, args),
**distribute_utils.select_replica(
index, kwargs)))
else:
return_values.append(
fn(
*distribute_utils.select_replica(
index, args),
**distribute_utils.select_replica(
index, kwargs)))
finally:
_replica_index.graph_outside_run = None
_replica_index.current = None
return distribute_utils.regroup(return_values)
def _make_per_replica(values, devices, regroup=False):
devices = _get_devices(devices)
assert len(values) == len(devices)
# We simulate the result of regroup called on PerReplica which strips the
# PerReplica wrapper if it has only one value.
if len(values) == 1 and regroup:
with ops.device(devices[0]):
placed_v = array_ops.identity(values[0])
return placed_v
index = []
for d, v in zip(devices, values):
with ops.device(d):
placed_v = array_ops.identity(v)
index.append(placed_v)
return distribute_utils.regroup(index)
def reduce(self,
reduce_op,
per_replica_value,
destinations,
experimental_hints=None):
"""Reduce `per_replica_value` to `destinations`.
It runs the reduction operation defined by `reduce_op` and put the
result on `destinations`.
Args:
reduce_op: An instance of `tf.distribute.ReduceOp` that indicates how
per_replica_value will be reduced.
per_replica_value: A `tf.distribute.DistributedValues` object or a tensor
with device set.
destinations: the reduction destinations.
experimental_hints: A `tf.distrbute.experimental.CollectiveHints`. Hints
to perform collective operations.
Returns:
a Mirrored object.
Raises:
ValueError: if per_replica_value can't be converted to a PerReplica
object or if destinations aren't strings, Variables or DistributedValues
"""
if not isinstance(per_replica_value, value_lib.DistributedValues):
per_replica_value = _make_tensor_into_per_replica(
per_replica_value)
validate_destinations(destinations)
# Shortcut if `per_replica_value` only contains one value.
if self._num_between_graph_workers == 1 and len(
per_replica_value.values) == 1 and _devices_match(
per_replica_value, destinations):
with ops.device(per_replica_value.values[0].device):
v = array_ops.identity(per_replica_value.values[0])
return distribute_utils.regroup((v, ),
wrap_class=value_lib.Mirrored)
if experimental_hints is None:
experimental_hints = collective_util.Hints()
return self.reduce_implementation(reduce_op, per_replica_value,
destinations, experimental_hints)
def testSameId(self):
foo = object()
result = distribute_utils.regroup((("a", foo), ("b", foo)))
self.assertIsInstance(result, tuple)
self.assertLen(result, 2)
self._is_per_replica(result[0], ["a", "b"])
self.assertIs(foo, result[1])
# Test select_replica(), should undo the merge done by regroup().
result_0 = distribute_utils.select_replica(0, result)
self.assertIsInstance(result_0, tuple)
self.assertLen(result_0, 2)
self.assertEqual("a", result_0[0])
self.assertIs(foo, result_0[1])
result_1 = distribute_utils.select_replica(1, result)
self.assertIsInstance(result_1, tuple)
self.assertLen(result_1, 2)
self.assertEqual("b", result_1[0])
self.assertIs(foo, result_1[1])
def testRegroupCollectionsMapping(self):
class CollectionsMappingBasedClass(collections.Mapping):
"""Class inherited from collections.Mapping."""
def __init__(self, *args, **kwargs):
self._d = dict(*args, **kwargs)
def __getitem__(self, key):
return self._d.__getitem__(key)
def __iter__(self):
return iter(self._d)
def __len__(self):
return len(self._d)
result = distribute_utils.regroup(
(CollectionsMappingBasedClass(a="a1", b="b1"),
CollectionsMappingBasedClass(a="a2", b="b2")))
self.assertIsInstance(result, CollectionsMappingBasedClass)
self._is_per_replica(result["a"], ["a1", "a2"])
self._is_per_replica(result["b"], ["b1", "b2"])
def _do_batch_all_reduce_sparse(self, reduce_op, per_replica_values,
experimental_hints):
"""All-reduce IndexedSlices across all workers in a batch."""
logging.log_first_n(
logging.INFO, "Collective batch_all_reduce for IndexedSlices: "
"%d all-reduces, group_size = %d" %
(len(per_replica_values), self._group_size), 10)
# Pass self._communication to the runtime as a communication hint.
communication_hint = self._communication.value
# For now, we use NCCL only when batch_size > 1.
# TODO(b/132575814): switch to NCCL for all collectives when communication
# is NCCL.
if self._communication == CollectiveCommunication.NCCL and len(
per_replica_values) == 1:
communication_hint = CollectiveCommunication.AUTO.value
gathered_values = []
with self._lock, ops.name_scope("allreduce"):
for per_replica in per_replica_values:
gathered_values.append(
cross_device_utils.build_collective_gather_indexed_slices(
per_replica.values,
self._devices,
self._group_size,
self._collective_keys,
communication_hint,
timeout=experimental_hints.timeout_seconds))
mirrored = []
for value in gathered_values:
if reduce_op == reduce_util.ReduceOp.MEAN:
# Assume each worker has the same number of replicas.
for i, v in enumerate(value):
with ops.device(v.device):
value[i].values = value[i].values / self._group_size
mirrored.append(
distribute_utils.regroup(value, wrap_class=value_lib.Mirrored))
return mirrored
def _ungroup_and_make_mirrored(grouped_reduced,
destinations,
reduce_op,
num_between_graph_workers=1):
"""Ungroup results from all-reduce and make Mirrored objects.
Each all-reduce result will be divided by the number of destinations before
Mirrored objects are created if reduce_op is "mean".
Args:
grouped_reduced: a list of lists, each sublist has components for each
device, paired with a None. It is the result from
cross_device_utils.aggregate_gradients_using*.
destinations: a value to colocate the result with.
reduce_op: Indicates how values will be aggregated. Accepted values
are `tf.distribute.ReduceOp.SUM`, `tf.distribute.ReduceOp.MEAN`.
num_between_graph_workers: number of workers in the between-graph
replication.
Returns:
a list of Mirrored objects.
"""
num_replicas = len(
get_devices_from(destinations)) * num_between_graph_workers
index = [[] for _ in range(len(grouped_reduced[0]))]
for per_replica_reduced in grouped_reduced:
for i, (v, _) in enumerate(per_replica_reduced):
if reduce_op == reduce_util.ReduceOp.MEAN:
with ops.device(v.device):
index[i].append(v / num_replicas)
else:
index[i].append(v)
return [
distribute_utils.regroup(v, wrap_class=value_lib.Mirrored)
for v in index
]
def _call_for_each_replica(distribution, fn, args, kwargs):
"""Run `fn` in separate threads, once per replica/worker device.
Args:
distribution: the DistributionStrategy object.
fn: function to run (will be run once per replica, each in its own thread).
args: positional arguments for `fn`
kwargs: keyword arguments for `fn`.
Returns:
Merged return value of `fn` across all replicas.
Raises:
RuntimeError: If fn() calls get_replica_context().merge_call() a different
number of times from the available devices.
"""
# TODO(josh11b): Add this option once we add synchronization to variable
# creation. Until then, this is pretty unsafe to use.
run_concurrently = False
if not context.executing_eagerly():
# Needed for per-thread device, etc. contexts in graph mode.
ops.get_default_graph().switch_to_thread_local()
coord = coordinator.Coordinator(
clean_stop_exception_types=(_RequestedStop, ))
shared_variable_store = {}
devices = distribution.extended.worker_devices
thread_local_callables = _get_thread_local_configuration_callable()
# TODO(isaprykin): Create these threads once instead of during every call.
threads = []
for index in range(len(devices)):
variable_creator_fn = shared_variable_creator.make_fn(
shared_variable_store, index)
t = _MirroredReplicaThread(
distribution, coord, index, devices, variable_creator_fn, fn,
distribute_utils.caching_scope_local,
distribute_utils.select_replica(index, args),
distribute_utils.select_replica(index,
kwargs), thread_local_callables)
threads.append(t)
for t in threads:
t.start()
# When `fn` starts `should_run` event is set on _MirroredReplicaThread
# (`MRT`) threads. The execution waits until
# `MRT.has_paused` is set, which indicates that either `fn` is
# complete or a `get_replica_context().merge_call()` is called. If `fn` is
# complete, then `MRT.done` is set to True. Otherwise, arguments
# of `get_replica_context().merge_call` from all paused threads are grouped
# and the `merge_fn` is performed. Results of the
# `get_replica_context().merge_call` are then set to `MRT.merge_result`.
# Each such `get_replica_context().merge_call` call returns the
# `MRT.merge_result` for that thread when `MRT.should_run` event
# is reset again. Execution of `fn` resumes.
try:
with coord.stop_on_exception():
all_done = False
while not all_done and not coord.should_stop():
done = []
if run_concurrently:
for t in threads:
t.should_run.set()
for t in threads:
t.has_paused.wait()
t.has_paused.clear()
if coord.should_stop():
return None
done.append(t.done)
else:
for t in threads:
t.should_run.set()
t.has_paused.wait()
t.has_paused.clear()
if coord.should_stop():
return None
done.append(t.done)
if coord.should_stop():
return None
all_done = all(done)
if not all_done:
if any(done):
raise RuntimeError(
"Some replicas made a different number of "
"replica_context().merge_call() calls.")
# get_replica_context().merge_call() case
merge_args = distribute_utils.regroup(
tuple(t.merge_args for t in threads))
merge_kwargs = distribute_utils.regroup(
tuple(t.merge_kwargs for t in threads))
# We capture the name_scope of the MRT when we call merge_fn
# to ensure that if we have opened a name scope in the MRT,
# it will be respected when executing the merge function. We only
# capture the name_scope from the first MRT and assume it is
# the same for all other MRTs.
mtt_captured_name_scope = threads[0].captured_name_scope
mtt_captured_var_scope = threads[0].captured_var_scope
# Capture and merge the control dependencies from all the threads.
mtt_captured_control_deps = set()
for t in threads:
mtt_captured_control_deps.update(
t.captured_control_deps)
with ops.name_scope(mtt_captured_name_scope),\
ops.control_dependencies(mtt_captured_control_deps), \
variable_scope.variable_scope(mtt_captured_var_scope):
merge_result = threads[0].merge_fn(
distribution, *merge_args, **merge_kwargs)
for r, t in enumerate(threads):
t.merge_result = distribute_utils.select_replica(
r, merge_result)
finally:
for t in threads:
t.should_run.set()
coord.join(threads)
return distribute_utils.regroup(tuple(t.main_result for t in threads))
def _experimental_run_steps_on_iterator(self,
fn,
iterator,
iterations,
initial_loop_values=None):
if initial_loop_values is None:
initial_loop_values = {}
initial_loop_values = nest.flatten(initial_loop_values)
ctx = input_lib.MultiStepContext()
def body(i, *args):
"""A wrapper around `fn` to create the while loop body."""
del args
fn_result = fn(ctx, iterator.get_next())
for (name, output) in ctx.last_step_outputs.items():
# Convert all outputs to tensors, potentially from `DistributedValues`.
ctx.last_step_outputs[name] = self._local_results(output)
flat_last_step_outputs = nest.flatten(ctx.last_step_outputs)
with ops.control_dependencies([fn_result]):
return [i + 1] + flat_last_step_outputs
# We capture the control_flow_context at this point, before we run `fn`
# inside a while_loop. This is useful in cases where we might need to exit
# these contexts and get back to the outer context to do some things, for
# e.g. create an op which should be evaluated only once at the end of the
# loop on the host. One such usage is in creating metrics' value op.
self._outer_control_flow_context = (
ops.get_default_graph()._get_control_flow_context()) # pylint: disable=protected-access
cond = lambda i, *args: i < iterations
i = constant_op.constant(0)
loop_result = control_flow_ops.while_loop(cond,
body,
[i] + initial_loop_values,
name="",
parallel_iterations=1,
back_prop=False,
swap_memory=False,
return_same_structure=True)
del self._outer_control_flow_context
ctx.run_op = control_flow_ops.group(loop_result)
# Convert the last_step_outputs from a list to the original dict structure
# of last_step_outputs.
last_step_tensor_outputs = loop_result[1:]
last_step_tensor_outputs_dict = nest.pack_sequence_as(
ctx.last_step_outputs, last_step_tensor_outputs)
for name, reduce_op in ctx._last_step_outputs_reduce_ops.items(): # pylint: disable=protected-access
output = last_step_tensor_outputs_dict[name]
# For outputs that have already been reduced, wrap them in a Mirrored
# container, else in a PerReplica container.
if reduce_op is None:
last_step_tensor_outputs_dict[name] = distribute_utils.regroup(
output)
else:
assert len(output) == 1
last_step_tensor_outputs_dict[name] = output[0]
ctx._set_last_step_outputs(last_step_tensor_outputs_dict) # pylint: disable=protected-access
return ctx
def testWrapAListOfTwoTuples(self):
result = distribute_utils.regroup([("1", "2"), ("3", "4")])
self.assertIsInstance(result, tuple)
self.assertLen(result, 2)
self._is_per_replica(result[0], ("1", "3"), values.PerReplica)
self._is_per_replica(result[1], ("2", "4"), values.PerReplica)
def _make_mirrored_indexed_slices(devices, values, indices, dense_shape):
values = [_make_indexed_slices(values, indices, dense_shape, d)
for d in devices]
return distribute_utils.regroup(
values,
wrap_class=value_lib.Mirrored)
def _call_for_each_replica(self, fn, args, kwargs):
with distribute_lib.ReplicaContext(
self._container_strategy(),
replica_id_in_sync_group=constant_op.constant(0, dtypes.int32)):
# TODO(rchao): Support multi-replica per worker or sync-group.
return distribute_utils.regroup((fn(*args, **kwargs),))
def tpu_function(args, kwargs):
"""TF Function used to replicate the user computation."""
if kwargs is None:
kwargs = {}
# Remove None at the end of args as they are not replicatable
# If there are None in the middle we can't do anything about it
# so let those cases fail.
# For example when Keras model predict is used they pass the targets as
# None. We want to handle it here so all client libraries don't have to
# do this as other strategies can handle None values better.
while args and args[-1] is None:
args = args[:-1]
# Used to re-structure flattened output tensors from `tpu.replicate()`
# into a structured format.
result = [[]]
def replicated_fn(replica_id, replica_args, replica_kwargs):
"""Wraps user function to provide replica ID and `Tensor` inputs."""
with _TPUReplicaContext(strategy,
replica_id_in_sync_group=replica_id):
result[0] = fn(*replica_args, **replica_kwargs)
return result[0]
replicate_inputs = [] # By replica.
for i in range(strategy.num_replicas_in_sync):
replicate_inputs.append([
constant_op.constant(i, dtype=dtypes.int32),
distribute_utils.select_replica(i, args),
distribute_utils.select_replica(i, kwargs)
])
# Construct and pass `maximum_shapes` so that we could support dynamic
# shapes using dynamic padder.
if options.experimental_enable_dynamic_batch_size and replicate_inputs:
maximum_shapes = []
flattened_list = nest.flatten(replicate_inputs[0])
for input_tensor in flattened_list:
if tensor_util.is_tensor(input_tensor):
rank = input_tensor.shape.rank
else:
rank = np.ndim(input_tensor)
maximum_shape = tensor_shape.TensorShape([None] * rank)
maximum_shapes.append(maximum_shape)
maximum_shapes = nest.pack_sequence_as(replicate_inputs[0],
maximum_shapes)
else:
maximum_shapes = None
if options.experimental_bucketizing_dynamic_shape:
padding_spec = tpu.PaddingSpec.POWER_OF_TWO
else:
padding_spec = None
with strategy.scope():
replicate_outputs = tpu.replicate(
replicated_fn,
replicate_inputs,
device_assignment=self._device_assignment,
maximum_shapes=maximum_shapes,
padding_spec=padding_spec,
xla_options=tpu.XLAOptions(
use_spmd_for_xla_partitioning=False))
# Remove all no ops that may have been added during 'tpu.replicate()'
if isinstance(result[0], list):
result[0] = [
output for output in result[0]
if not isinstance(output, ops.Operation)
]
# Workaround for `tpu.replicate` behaviour when single `Tensor` returned.
if result[0] is None or isinstance(result[0], ops.Operation):
replicate_outputs = [None] * len(replicate_outputs)
else:
replicate_outputs = [
nest.pack_sequence_as(result[0],
nest.flatten(replica_output))
for replica_output in replicate_outputs
]
return distribute_utils.regroup(replicate_outputs)
def testOneDevice(self):
result = distribute_utils.regroup((_nested_value("1"), ))
# On one device regroup() and select_replica() are basically identity.
self.assertEqual(_nested_value("1"), result)
self.assertEqual(_nested_value("1"),
distribute_utils.select_replica(0, result))
def _do_batch_all_reduce_dense(self, reduce_op, per_replica_values,
experimental_hints):
"""All-reduce across all workers in a batch."""
batch_size = len(per_replica_values)
# Pass self._communication to the runtime as a communication hint.
communication = self._communication.value
# For now, we use NCCL only when batch_size > 1.
# TODO(b/132575814): switch to NCCL for all collectives when communication
# is NCCL.
if self._communication == CollectiveCommunication.NCCL and batch_size == 1:
communication = CollectiveCommunication.AUTO.value
# Reverse the lists so that there's better chance that values follows
# the order in which they are calculated (e.g. when they're gradients), so
# as to overlap calculation with communication. However, this may not be
# optimal for cases like gradients of complicated non-sequential models.
#
# Note that we reverse the list before packing so that the first pack won't
# be too small, since it's more likely for first few packs to have long
# queuing time due to concurrent intense computation.
#
# TODO(b/147393503): explore solutions for optimal ordering.
packs = cross_device_utils.pack_by_size(
list(reversed(per_replica_values)),
experimental_hints.bytes_per_pack)
if batch_size > 1:
logging.info(
"Collective batch_all_reduce: %d all-reduces, num_devices = %d, "
"group_size = %d, communication_hint = %s, num_packs = %d",
batch_size, len(self._devices), self._group_size,
communication, len(packs))
else:
logging.log_first_n(
logging.INFO, "Collective batch_all_reduce: %d all-reduces, "
"num_devices = %d, group_size = %d, communication_hint = %s, "
"num_packs = %d" %
(batch_size, len(self._devices), self._group_size,
communication, len(packs)), 10)
reduced_values = []
with self._lock:
for pack in packs:
# By placing all CollectiveReduce ops in a pack under single name scope,
# we ensure they will be picked up by the `ScopedAllocator` grappler
# optimizer and packed into a single all-reduce.
with ops.name_scope("allreduce"):
for per_replica in pack:
# Add control dependencies per device from the last gradients to the
# current set, in order to serialize NCCL launches.
if (communication == CollectiveCommunication.NCCL.value
and reduced_values):
control_inputs = list(reduced_values[-1])
else:
control_inputs = None
reduced_values.append(
cross_device_utils.build_collective_reduce(
per_replica.values,
self._devices,
self._group_size,
self._collective_keys,
"Add",
"Id",
communication,
control_inputs,
executors=self._executors,
timeout=experimental_hints.timeout_seconds))
for e in self._executors:
e.wait()
mirrored = []
# Reverse the order of reduced value to recover the order in the input.
for value in reversed(reduced_values):
if reduce_op == reduce_util.ReduceOp.MEAN:
for i, v in enumerate(value):
with ops.device(v.device):
value[i] = v / self._group_size
mirrored.append(
distribute_utils.regroup(value, wrap_class=value_lib.Mirrored))
return mirrored
