Function bins_to_scalar 

pub fn bins_to_scalar(
    logits: &Tensor,
    min_val: f64,
    max_val: f64,
) -> Result<Tensor>

Convert bin logits to a scalar via softmax + weighted average of bin centres.

logits — [..., n_bins] min_val — value of the first bin centre max_val — value of the last bin centre

Returns [...] with the same leading dims, values in [min_val, max_val].