tf.distribute.experimental.MultiWorkerMirroredStrategy

A distribution strategy for synchronous training on multiple workers.

Inherits From: Strategy

tf.distribute.experimental.MultiWorkerMirroredStrategy(
    communication=tf.distribute.experimental.CollectiveCommunication.AUTO,
    cluster_resolver=None
)

This strategy implements synchronous distributed training across multiple workers, each with potentially multiple GPUs. Similar to tf.distribute.MirroredStrategy, it creates copies of all variables in the model on each device across all workers.

It uses CollectiveOps's implementation of multi-worker all-reduce to to keep variables in sync. A collective op is a single op in the TensorFlow graph which can automatically choose an all-reduce algorithm in the TensorFlow runtime according to hardware, network topology and tensor sizes.

By default it uses all local GPUs or CPU for single-worker training.

When 'TF_CONFIG' environment variable is set, it parses cluster_spec, task_type and task_id from 'TF_CONFIG' and turns into a multi-worker strategy which mirrored models on GPUs of all machines in a cluster. In the current implementation, it uses all GPUs in a cluster and it assumes all workers have the same number of GPUs.

You can also pass a distribute.cluster_resolver.ClusterResolver instance when instantiating the strategy. The task_type, task_id etc. will be parsed from the resolver instance instead of from the TF_CONFIG env var.

It supports both eager mode and graph mode. However, for eager mode, it has to set up the eager context in its constructor and therefore all ops in eager mode have to run after the strategy object is created.

Args
communication	optional Enum of type distribute.experimental.CollectiveCommunication. This provides a way for the user to override the choice of collective op communication. Possible values include AUTO, RING, and NCCL.
cluster_resolver	optional distribute.cluster_resolver.ClusterResolver object. The default ClusterResolver that is used is the TFConfigClusterResolver which is instantiated from the TF_CONFIG env var.

Attributes
cluster_resolver	Returns the cluster resolver associated with this strategy. As a multi-worker strategy, tf.distribute.experimental.MultiWorkerMirroredStrategy provides the associated tf.distribute.cluster_resolver.ClusterResolver. If the user provides one in __init__, that instance is returned; if the user does not, a default TFConfigClusterResolver is provided.
extended	tf.distribute.StrategyExtended with additional methods.
num_replicas_in_sync	Returns number of replicas over which gradients are aggregated.

Methods

experimental_assign_to_logical_device

View source

experimental_assign_to_logical_device(
    tensor, logical_device_id
)

Adds annotation that tensor will be assigned to a logical device.

Note: This API is only supported in TPUStrategy for now. This adds an annotation to tensor specifying that operations on tensor will be invoked on logical core device id logical_device_id. When model parallelism is used, the default behavior is that all ops are placed on zero-th logical device.

# Initializing TPU system with 2 logical devices and 4 replicas.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(tpu='')
tf.config.experimental_connect_to_cluster(resolver)
topology = tf.tpu.experimental.initialize_tpu_system(resolver)
device_assignment = tf.tpu.experimental.DeviceAssignment.build(
    topology,
    computation_shape=[1, 1, 1, 2],
    num_replicas=4)
strategy = tf.distribute.TPUStrategy(
    resolver, experimental_device_assignment=device_assignment)
iterator = iter(inputs)

@tf.function()
def step_fn(inputs):
  output = tf.add(inputs, inputs)

  # Add operation will be executed on logical device 0.
  output = strategy.experimental_assign_to_logical_device(output, 0)
  return output

strategy.run(step_fn, args=(next(iterator),))

Args
tensor	Input tensor to annotate.
logical_device_id	Id of the logical core to which the tensor will be assigned.

Raises
ValueError	The logical device id presented is not consistent with total number of partitions specified by the device assignment.

Returns
Annotated tensor with idential value as tensor.

experimental_distribute_dataset

View source

experimental_distribute_dataset(
    dataset, options=None
)

Creates tf.distribute.DistributedDataset from tf.data.Dataset.

The returned tf.distribute.DistributedDataset can be iterated over similar to how regular datasets can. NOTE: The user cannot add any more transformations to a tf.distribute.DistributedDataset.

The following is an example:

strategy = tf.distribute.MirroredStrategy()

# Create a dataset
dataset = dataset_ops.Dataset.TFRecordDataset([
  "/a/1.tfr", "/a/2.tfr", "/a/3.tfr", "/a/4.tfr"])

# Distribute that dataset
dist_dataset = strategy.experimental_distribute_dataset(dataset)

# Iterate over the `tf.distribute.DistributedDataset`
for x in dist_dataset:
  # process dataset elements
  strategy.run(replica_fn, args=(x,))

In the code snippet above, the tf.distribute.DistributedDataset dist_dataset is batched by GLOBAL_BATCH_SIZE, and we iterate through it using for x in dist_dataset. x a tf.distribute.DistributedValues containing data for all replicas, which aggregates to a batch of GLOBAL_BATCH_SIZE. tf.distribute.Strategy.run will take care of feeding the right per-replica data in x to the right replica_fn executed on each replica.

What's under the hood of this method, when we say the tf.data.Dataset instance - dataset - gets distributed? It depends on how you set the tf.data.experimental.AutoShardPolicy through tf.data.experimental.DistributeOptions. By default, it is set to tf.data.experimental.AutoShardPolicy.AUTO. In a multi-worker setting, we will first attempt to distribute dataset by detecting whether dataset is being created out of reader datasets (e.g. tf.data.TFRecordDataset, tf.data.TextLineDataset, etc.) and if so, try to shard the input files. Note that there has to be at least one input file per worker. If you have less than one input file per worker, we suggest that you disable dataset sharding across workers, by setting the tf.data.experimental.DistributeOptions.auto_shard_policy to be tf.data.experimental.AutoShardPolicy.OFF.

If the attempt to shard by file is unsuccessful (i.e. the dataset is not read from files), we will shard the dataset evenly at the end by appending a .shard operation to the end of the processing pipeline. This will cause the entire preprocessing pipeline for all the data to be run on every worker, and each worker will do redundant work. We will print a warning if this route is selected.

As mentioned before, within each worker, we will also split the data among all the worker devices (if more than one a present). This will happen even if multi-worker sharding is disabled.

If the above batch splitting and dataset sharding logic is undesirable, please use tf.distribute.Strategy.experimental_distribute_datasets_from_function instead, which does not do any automatic splitting or sharding.

You can also use the element_spec property of the tf.distribute.DistributedDataset instance returned by this API to query the tf.TypeSpec of the elements returned by the iterator. This can be used to set the input_signature property of a tf.function.

strategy = tf.distribute.MirroredStrategy()

# Create a dataset
dataset = dataset_ops.Dataset.TFRecordDataset([
  "/a/1.tfr", "/a/2.tfr", "/a/3.tfr", "/a/4.tfr"])

# Distribute that dataset
dist_dataset = strategy.experimental_distribute_dataset(dataset)

@tf.function(input_signature=[dist_dataset.element_spec])
def train_step(inputs):
  # train model with inputs
  return

# Iterate over the `tf.distribute.DistributedDataset`
for x in dist_dataset:
  # process dataset elements
  strategy.run(train_step, args=(x,))

Note: The order in which the data is processed by the workers when using tf.distribute.Strategy.experimental_distribute_dataset or tf.distribute.Strategy.experimental_distribute_datasets_from_function is not guaranteed. This is typically required if you are using tf.distribute to scale prediction. You can however insert an index for each element in the batch and order outputs accordingly. Refer to this snippet for an example of how to order outputs.

Args
dataset	tf.data.Dataset that will be sharded across all replicas using the rules stated above.
options	tf.distribute.InputOptions used to control options on how this dataset is distributed.

Returns
A tf.distribute.DistributedDataset.

experimental_distribute_datasets_from_function

View source

experimental_distribute_datasets_from_function(
    dataset_fn, options=None
)

Distributes tf.data.Dataset instances created by calls to dataset_fn.

dataset_fn will be called once for each worker in the strategy. Each replica on that worker will dequeue one batch of inputs from the local Dataset (i.e. if a worker has two replicas, two batches will be dequeued from the Dataset every step).

This method can be used for several purposes. For example, where experimental_distribute_dataset is unable to shard the input files, this method might be used to manually shard the dataset (avoiding the slow fallback behavior in experimental_distribute_dataset). In cases where the dataset is infinite, this sharding can be done by creating dataset replicas that differ only in their random seed. experimental_distribute_dataset may also sometimes fail to split the batch across replicas on a worker. In that case, this method can be used where that limitation does not exist.

The dataset_fn should take an tf.distribute.InputContext instance where information about batching and input replication can be accessed.

You can also use the element_spec property of the tf.distribute.DistributedDataset returned by this API to query the tf.TypeSpec of the elements returned by the iterator. This can be used to set the input_signature property of a tf.function.

global_batch_size = 8
def dataset_fn(input_context):
  batch_size = input_context.get_per_replica_batch_size(
                   global_batch_size)
  d = tf.data.Dataset.from_tensors([[1.]]).repeat().batch(batch_size)
  return d.shard(
      input_context.num_input_pipelines,
      input_context.input_pipeline_id)

strategy = tf.distribute.MirroredStrategy()
ds = strategy.experimental_distribute_datasets_from_function(dataset_fn)

def train(ds):
  @tf.function(input_signature=[ds.element_spec])
  def step_fn(inputs):
    # train the model with inputs
    return inputs

... for batch in ds: ... replica_results = strategy.run(replica_fn, args=(batch,))

train(ds)

Key Point: The tf.data.Dataset returned by dataset_fn should have a per-replica batch size, unlike experimental_distribute_dataset, which uses the global batch size. This may be computed using input_context.get_per_replica_batch_size.

Note: The order in which the data is processed by the workers when using tf.distribute.Strategy.experimental_distribute_dataset or tf.distribute.Strategy.experimental_distribute_datasets_from_function is not guaranteed. This is typically required if you are using tf.distribute to scale prediction. You can however insert an index for each element in the batch and order outputs accordingly. Refer to this snippet for an example of how to order outputs.

Args
dataset_fn	A function taking a tf.distribute.InputContext instance and returning a tf.data.Dataset.
options	tf.distribute.InputOptions used to control options on how this dataset is distributed.

Returns
A tf.distribute.DistributedDataset.

experimental_distribute_values_from_function

View source

experimental_distribute_values_from_function(
    value_fn
)

Generates tf.distribute.DistributedValues from value_fn.

This function is to generate tf.distribute.DistributedValues to pass into run, reduce, or other methods that take distributed values when not using datasets.

Args
value_fn	The function to run to generate values. It is called for each replica with tf.distribute.ValueContext as the sole argument. It must return a Tensor or a type that can be converted to a Tensor.

Returns
A tf.distribute.DistributedValues containing a value for each replica.

Example usage:

1. Return constant value per replica:

strategy = tf.distribute.MirroredStrategy()
def value_fn(ctx):
  return tf.constant(1.)
distributed_values = (
     strategy.experimental_distribute_values_from_function(
       value_fn))
local_result = strategy.experimental_local_results(distributed_values)
local_result
(<tf.Tensor: shape=(), dtype=float32, numpy=1.0>,)

1. Distribute values in array based on replica_id:

strategy = tf.distribute.MirroredStrategy()
array_value = np.array([3., 2., 1.])
def value_fn(ctx):
  return array_value[ctx.replica_id_in_sync_group]
distributed_values = (
     strategy.experimental_distribute_values_from_function(
       value_fn))
local_result = strategy.experimental_local_results(distributed_values)
local_result
(3.0,)

1. Specify values using num_replicas_in_sync:

strategy = tf.distribute.MirroredStrategy()
def value_fn(ctx):
  return ctx.num_replicas_in_sync
distributed_values = (
     strategy.experimental_distribute_values_from_function(
       value_fn))
local_result = strategy.experimental_local_results(distributed_values)
local_result
(1,)

1. Place values on devices and distribute:

strategy = tf.distribute.TPUStrategy()
worker_devices = strategy.extended.worker_devices
multiple_values = []
for i in range(strategy.num_replicas_in_sync):
  with tf.device(worker_devices[i]):
    multiple_values.append(tf.constant(1.0))

def value_fn(ctx):
  return multiple_values[ctx.replica_id_in_sync_group]

distributed_values = strategy.
  experimental_distribute_values_from_function(
  value_fn)

experimental_local_results

View source

experimental_local_results(
    value
)

Returns the list of all local per-replica values contained in value.

Note: This only returns values on the worker initiated by this client. When using a tf.distribute.Strategy like tf.distribute.experimental.MultiWorkerMirroredStrategy, each worker will be its own client, and this function will only return values computed on that worker.

Args
value	A value returned by experimental_run(), run(), extended.call_for_each_replica(), or a variable created in scope.

Returns
A tuple of values contained in value. If value represents a single value, this returns (value,).

experimental_make_numpy_dataset

View source

experimental_make_numpy_dataset(
    numpy_input
)

Makes a tf.data.Dataset from a numpy array. (deprecated)

Warning: THIS FUNCTION IS DEPRECATED. It will be removed after 2020-09-30. Instructions for updating: Please use tf.data.Dataset.from_tensor_slices instead

This avoids adding numpy_input as a large constant in the graph, and copies the data to the machine or machines that will be processing the input.

Note that you will likely need to use experimental_distribute_dataset with the returned dataset to further distribute it with the strategy.

Example:

strategy = tf.distribute.MirroredStrategy()
numpy_input = np.ones([10], dtype=np.float32)
dataset = strategy.experimental_make_numpy_dataset(numpy_input)
dataset
<TensorSliceDataset shapes: (), types: tf.float32>
dataset = dataset.batch(2)
dist_dataset = strategy.experimental_distribute_dataset(dataset)

Args
numpy_input	a nest of NumPy input arrays that will be converted into a dataset. Note that the NumPy arrays are stacked, as that is normal tf.data.Dataset behavior.

Returns
A tf.data.Dataset representing numpy_input.

experimental_replicate_to_logical_devices

View source

experimental_replicate_to_logical_devices(
    tensor
)

Adds annotation that tensor will be replicated to all logical devices.

Note: This API is only supported in TPUStrategy for now. This adds an annotation to tensor tensor specifying that operations on tensor will be invoked on all logical devices.

# Initializing TPU system with 2 logical devices and 4 replicas.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(tpu='')
tf.config.experimental_connect_to_cluster(resolver)
topology = tf.tpu.experimental.initialize_tpu_system(resolver)
device_assignment = tf.tpu.experimental.DeviceAssignment.build(
    topology,
    computation_shape=[1, 1, 1, 2],
    num_replicas=4)
strategy = tf.distribute.TPUStrategy(
    resolver, experimental_device_assignment=device_assignment)

iterator = iter(inputs)

@tf.function()
def step_fn(inputs):
  images, labels = inputs
  images = strategy.experimental_split_to_logical_devices(
    inputs, [1, 2, 4, 1])

  # model() function will be executed on 8 logical devices with `inputs`
  # split 2 * 4  ways.
  output = model(inputs)

  # For loss calculation, all logical devices share the same logits
  # and labels.
  labels = strategy.experimental_replicate_to_logical_devices(labels)
  output = strategy.experimental_replicate_to_logical_devices(output)
  loss = loss_fn(labels, output)

  return loss

strategy.run(step_fn, args=(next(iterator),))

Args: tensor: Input tensor to annotate.

Returns
Annotated tensor with idential value as tensor.

experimental_split_to_logical_devices

View source

experimental_split_to_logical_devices(
    tensor, partition_dimensions
)

Adds annotation that tensor will be split across logical devices.

Note: This API is only supported in TPUStrategy for now. This adds an annotation to tensor tensor specifying that operations on tensor will be be split among multiple logical devices. Tensor tensor will be split across dimensions specified by partition_dimensions. The dimensions of tensor must be divisible by corresponding value in partition_dimensions.

For example, for system with 8 logical devices, if tensor is an image tensor with shape (batch_size, width, height, channel) and partition_dimensions is [1, 2, 4, 1], then tensor will be split 2 in width dimension and 4 way in height dimension and the split tensor values will be fed into 8 logical devices.

# Initializing TPU system with 8 logical devices and 1 replica.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(tpu='')
tf.config.experimental_connect_to_cluster(resolver)
topology = tf.tpu.experimental.initialize_tpu_system(resolver)
device_assignment = tf.tpu.experimental.DeviceAssignment.build(
    topology,
    computation_shape=[1, 2, 2, 2],
    num_replicas=1)
strategy = tf.distribute.TPUStrategy(
    resolver, experimental_device_assignment=device_assignment)

iterator = iter(inputs)

@tf.function()
def step_fn(inputs):
  inputs = strategy.experimental_split_to_logical_devices(
    inputs, [1, 2, 4, 1])

  # model() function will be executed on 8 logical devices with `inputs`
  # split 2 * 4  ways.
  output = model(inputs)
  return output

strategy.run(step_fn, args=(next(iterator),))

Args: tensor: Input tensor to annotate. partition_dimensions: An unnested list of integers with the size equal to rank of tensor specifying how tensor will be partitioned. The product of all elements in partition_dimensions must be equal to the total number of logical devices per replica.

Raises
ValueError	1) If the size of partition_dimensions does not equal to rank of tensor or 2) if product of elements of partition_dimensions does not match the number of logical devices per replica defined by the implementing DistributionStrategy's device specification or 3) if a known size of tensor is not divisible by corresponding value in partition_dimensions.

Returns
Annotated tensor with idential value as tensor.

reduce

View source

reduce(
    reduce_op, value, axis
)

Reduce value across replicas.

Given a per-replica value returned by run, say a per-example loss, the batch will be divided across all the replicas. This function allows you to aggregate across replicas and optionally also across batch elements. For example, if you have a global batch size of 8 and 2 replicas, values for examples [0, 1, 2, 3] will be on replica 0 and [4, 5, 6, 7] will be on replica 1. By default, reduce will just aggregate across replicas, returning [0+4, 1+5, 2+6, 3+7]. This is useful when each replica is computing a scalar or some other value that doesn't have a "batch" dimension (like a gradient). More often you will want to aggregate across the global batch, which you can get by specifying the batch dimension as the axis, typically axis=0. In this case it would return a scalar 0+1+2+3+4+5+6+7.

If there is a last partial batch, you will need to specify an axis so that the resulting shape is consistent across replicas. So if the last batch has size 6 and it is divided into [0, 1, 2, 3] and [4, 5], you would get a shape mismatch unless you specify axis=0. If you specify tf.distribute.ReduceOp.MEAN, using axis=0 will use the correct denominator of 6. Contrast this with computing reduce_mean to get a scalar value on each replica and this function to average those means, which will weigh some values 1/8 and others 1/4.

Args
reduce_op	A tf.distribute.ReduceOp value specifying how values should be combined.
value	A "per replica" value, e.g. returned by run to be combined into a single tensor.
axis	Specifies the dimension to reduce along within each replica's tensor. Should typically be set to the batch dimension, or None to only reduce across replicas (e.g. if the tensor has no batch dimension).

Returns
A Tensor.

run

View source

run(
    fn, args=(), kwargs=None, options=None
)

Run fn on each replica, with the given arguments.

Executes ops specified by fn on each replica. If args or kwargs have tf.distribute.DistributedValues, such as those produced by a tf.distribute.DistributedDataset from tf.distribute.Strategy.experimental_distribute_dataset or tf.distribute.Strategy.experimental_distribute_datasets_from_function, when fn is executed on a particular replica, it will be executed with the component of tf.distribute.DistributedValues that correspond to that replica.

fn may call tf.distribute.get_replica_context() to access members such as all_reduce.

All arguments in args or kwargs should either be nest of tensors or tf.distribute.DistributedValues containing tensors or composite tensors.

Key Point: Depending on the implementation of tf.distribute.Strategy and whether eager execution is enabled, fn may be called one or more times. If fn is annotated with tf.function or tf.distribute.Strategy.run is called inside a tf.function, eager execution is disabled and fn is called once (or once per replica, if you are using MirroredStrategy) to generate a Tensorflow graph, which will then be reused for execution with new inputs. Otherwise, if eager execution is enabled, fn will be called every step just like regular python code.

Example usage:

1. Constant tensor input.

strategy = tf.distribute.MirroredStrategy()
tensor_input = tf.constant(3.0)
@tf.function
def replica_fn(input):
  return input*2.0
result = strategy.run(replica_fn, args=(tensor_input,))
result
<tf.Tensor: shape=(), dtype=float32, numpy=6.0>

1. DistributedValues input.

strategy = tf.distribute.MirroredStrategy()
@tf.function
def run():
  def value_fn(value_context):
    return value_context.num_replicas_in_sync
  distributed_values = (
    strategy.experimental_distribute_values_from_function(
      value_fn))
  def replica_fn2(input):
    return input*2
  return strategy.run(replica_fn2, args=(distributed_values,))
result = run()
result
<tf.Tensor: shape=(), dtype=int32, numpy=2>

Args
fn	The function to run. The output must be a tf.nest of Tensors.
args	(Optional) Positional arguments to fn.
kwargs	(Optional) Keyword arguments to fn.
options	(Optional) An instance of tf.distribute.RunOptions specifying the options to run fn.

Returns
Merged return value of fn across replicas. The structure of the return value is the same as the return value from fn. Each element in the structure can either be tf.distribute.DistributedValues, Tensor objects, or Tensors (for example, if running on a single replica).

scope

View source

scope()

Returns a context manager selecting this Strategy as current.

Inside a with strategy.scope(): code block, this thread will use a variable creator set by strategy, and will enter its "cross-replica context".

In MultiWorkerMirroredStrategy, all variables created inside `strategy.scope() will be mirrored on all replicas of each worker. Moreover, it also sets a default device scope so that ops without specified devices will end up on the correct worker.

Returns
A context manager to use for creating variables with this strategy.