View source on GitHub |
A class with a collection of APIs that can be called in a replica context.
tf.compat.v1.distribute.ReplicaContext(
strategy, replica_id_in_sync_group
)
You can use tf.distribute.get_replica_context
to get an instance of
ReplicaContext
, which can only be called inside the function passed to
tf.distribute.Strategy.run
.
strategy = tf.distribute.MirroredStrategy(['GPU:0', 'GPU:1'])
def func():
replica_context = tf.distribute.get_replica_context()
return replica_context.replica_id_in_sync_group
strategy.run(func)
PerReplica:{
0: <tf.Tensor: shape=(), dtype=int32, numpy=0>,
1: <tf.Tensor: shape=(), dtype=int32, numpy=1>
}
Args | |
---|---|
strategy
|
A tf.distribute.Strategy .
|
replica_id_in_sync_group
|
An integer, a Tensor or None. Prefer an
integer whenever possible to avoid issues with nested tf.function . It
accepts a Tensor only to be compatible with tpu.replicate .
|
Attributes | |
---|---|
devices
|
Returns the devices this replica is to be executed on, as a tuple of strings. (deprecated) |
num_replicas_in_sync
|
Returns number of replicas that are kept in sync. |
replica_id_in_sync_group
|
Returns the id of the replica.
This identifies the replica among all replicas that are kept in sync. The
value of the replica id can range from 0 to
|
strategy
|
The current tf.distribute.Strategy object.
|
Methods
all_reduce
all_reduce(
reduce_op, value, options=None
)
All-reduces value
across all replicas.
strategy = tf.distribute.MirroredStrategy(["GPU:0", "GPU:1"])
def step_fn():
ctx = tf.distribute.get_replica_context()
value = tf.identity(1.)
return ctx.all_reduce(tf.distribute.ReduceOp.SUM, value)
strategy.experimental_local_results(strategy.run(step_fn))
(<tf.Tensor: shape=(), dtype=float32, numpy=2.0>,
<tf.Tensor: shape=(), dtype=float32, numpy=2.0>)
It supports batched operations. You can pass a list of values and it
attempts to batch them when possible. You can also specify options
to indicate the desired batching behavior, e.g. batch the values into
multiple packs so that they can better overlap with computations.
strategy = tf.distribute.MirroredStrategy(["GPU:0", "GPU:1"])
def step_fn():
ctx = tf.distribute.get_replica_context()
value1 = tf.identity(1.)
value2 = tf.identity(2.)
return ctx.all_reduce(tf.distribute.ReduceOp.SUM, [value1, value2])
strategy.experimental_local_results(strategy.run(step_fn))
([<tf.Tensor: shape=(), dtype=float32, numpy=2.0>,
<tf.Tensor: shape=(), dtype=float32, numpy=4.0>],
[<tf.Tensor: shape=(), dtype=float32, numpy=2.0>,
<tf.Tensor: shape=(), dtype=float32, numpy=4.0>])
Note that all replicas need to participate in the all-reduce, otherwise this operation hangs. Note that if there're multiple all-reduces, they need to execute in the same order on all replicas. Dispatching all-reduce based on conditions is usually error-prone.
Known limitation: if value
contains tf.IndexedSlices
, attempting to
compute gradient w.r.t value
would result in an error.
This API currently can only be called in the replica context. Other variants to reduce values across replicas are:
tf.distribute.StrategyExtended.reduce_to
: the reduce and all-reduce API in the cross-replica context.tf.distribute.StrategyExtended.batch_reduce_to
: the batched reduce and all-reduce API in the cross-replica context.tf.distribute.Strategy.reduce
: a more convenient method to reduce to the host in cross-replica context.
Args | |
---|---|
reduce_op
|
a tf.distribute.ReduceOp value specifying how values should
be combined. Allows using string representation of the enum such as
"SUM", "MEAN".
|
value
|
a potentially nested structure of tf.Tensor or tf.IndexedSlices which
tf.nest.flatten accepts. The structure and the shapes of value need to be
same on all replicas.
|
options
|
a tf.distribute.experimental.CommunicationOptions . Options to
perform collective operations. This overrides the default options if the
tf.distribute.Strategy takes one in the constructor. See
tf.distribute.experimental.CommunicationOptions for details of the
options.
|
Returns | |
---|---|
A nested structure of tf.Tensor with the reduced values. The structure
is the same as value .
|
merge_call
merge_call(
merge_fn, args=(), kwargs=None
)
Merge args across replicas and run merge_fn
in a cross-replica context.
This allows communication and coordination when there are multiple calls
to the step_fn triggered by a call to strategy.run(step_fn, ...)
.
See tf.distribute.Strategy.run
for an explanation.
If not inside a distributed scope, this is equivalent to:
strategy = tf.distribute.get_strategy()
with cross-replica-context(strategy):
return merge_fn(strategy, *args, **kwargs)
Args | |
---|---|
merge_fn
|
Function that joins arguments from threads that are given as
PerReplica. It accepts tf.distribute.Strategy object as
the first argument.
|
args
|
List or tuple with positional per-thread arguments for merge_fn .
|
kwargs
|
Dict with keyword per-thread arguments for merge_fn .
|
Returns | |
---|---|
The return value of merge_fn , except for PerReplica values which are
unpacked.
|