Connects N inputs to an N-way replicated TPU computation.

    inputs, is_mirrored_variable=False, index=-1, name=None

This operation holds a replicated input to a tpu.replicate() computation subgraph. Each replicated input has the same shape and type alongside the output.

For example:

%a = "tf.opA"()
%b = "tf.opB"()
%replicated_input = "tf.TPUReplicatedInput"(%a, %b)
%computation = "tf.Computation"(%replicated_input)

The above computation has a replicated input of two replicas.


  • inputs: A list of at least 1 Tensor objects with the same type.
  • is_mirrored_variable: An optional bool. Defaults to False.
  • index: An optional int. Defaults to -1.
  • name: A name for the operation (optional).


A Tensor. Has the same type as inputs.