tf.raw_ops.ConcatOffset

Computes offsets of concat inputs within its output.

tf.raw_ops.ConcatOffset(
    concat_dim, shape, name=None
)

For example:

x = [2, 2, 7]
y = [2, 3, 7]
z = [2, 9, 7]
offsets = concat_offset(1, [x, y, z])
[list(off.numpy()) for off in offsets]
[[0, 0, 0], [0, 2, 0], [0, 5, 0]]

This is typically used by gradient computations for a concat operation.

concat_dim A Tensor of type int32. The dimension along which to concatenate.
shape A list of at least 2 Tensor objects with the same type in: int32, int64. The N int32 or int64 vectors representing shape of tensors being concatenated.
name A name for the operation (optional).

A list with the same length as shape of Tensor objects with the same type as shape.