tf.raw_ops.ParallelConcat

Concatenates a list of N tensors along the first dimension.

tf.raw_ops.ParallelConcat(
    values, shape, name=None
)

The input tensors are all required to have size 1 in the first dimension.

For example:

# 'x' is [[1, 4]]
# 'y' is [[2, 5]]
# 'z' is [[3, 6]]
parallel_concat([x, y, z]) => [[1, 4], [2, 5], [3, 6]]  # Pack along first dim.

The difference between concat and parallel_concat is that concat requires all of the inputs be computed before the operation will begin but doesn't require that the input shapes be known during graph construction. Parallel concat will copy pieces of the input into the output as they become available, in some situations this can provide a performance benefit.

Args:

  • values: A list of at least 1 Tensor objects with the same type. Tensors to be concatenated. All must have size 1 in the first dimension and same shape.
  • shape: A tf.TensorShape or list of ints. the final shape of the result; should be equal to the shapes of any input but with the number of input values in the first dimension.
  • name: A name for the operation (optional).

Returns:

A Tensor. Has the same type as values.