tf.compat.v1.fixed_size_partitioner

Partitioner to specify a fixed number of shards along given axis.

tf.compat.v1.fixed_size_partitioner(
    num_shards, axis=0
)

Migrate to TF2

Caution: This API was designed for TensorFlow v1. Continue reading for details on how to migrate from this API to a native TensorFlow v2 equivalent. See the TensorFlow v1 to TensorFlow v2 migration guide for instructions on how to migrate the rest of your code.

This API is deprecated in TF2. In TF2, partitioner is no longer part of the variable declaration via tf.Variable. ParameterServer Training handles partitioning of variables. The corresponding TF2 partitioner class of fixed_size_partitioner is tf.distribute.experimental.partitioners.FixedShardsPartitioner.

Check the migration guide on the differences in treatment of variables and losses between TF1 and TF2.

Before:

  x = tf.compat.v1.get_variable(
    "x", shape=(2,), partitioner=tf.compat.v1.fixed_size_partitioner(2)
  )

After:

  partitioner = (
      tf.distribute.experimental.partitioners.FixedShardsPartitioner(
          num_shards=2)
  )
  strategy = tf.distribute.experimental.ParameterServerStrategy(
                 cluster_resolver=cluster_resolver,
                 variable_partitioner=partitioner)

  with strategy.scope():
    x = tf.Variable([1.0, 2.0])

Description

Args
num_shards	int, number of shards to partition variable.
axis	int, axis to partition on.

Returns
A partition function usable as the partitioner argument to variable_scope and get_variable.