durations (np.ndarray (1D, real)) – The durations \(\{\delta t_n\}\) of the \(N\) constant
segments.
values (np.ndarray or Tensor) – The values \(\{v_n\}\) of the function on the constant segments.
The dimension corresponding to time_dimension must be the same
length as durations. To create a batch of
\(B_1 \times \ldots \times B_n\) piecewise-constant tensors of
shape \(D_1 \times \ldots \times D_m\), provide this values
parameter as an object of shape
\(B_1\times\ldots\times B_n\times N\times D_1\times\ldots\times D_m\).
time_dimension (int, optional) – The axis along values corresponding to time. All dimensions that
come before the time_dimension are batch dimensions: if there are
\(n\) batch dimensions, then time_dimension is also \(n\).
Defaults to 0, which corresponds to no batch. Note that you can
pass a negative value to refer to the time dimension.
name (str, optional) – The name of the node.