INNER CODE UNIT · Python

means

nv-tlabs/ATISS · scene_synthesis/losses/__init__.py:63

    means = pred[:, :, nr_mix:2 * nr_mix]
    log_scales = torch.clamp(
        pred[:, :, 2 * nr_mix:3 * nr_mix], min=log_scale_min
    )

    centered_y = target - means
    inv_stdv = torch.exp(-log_scales)
    plus_in = inv_stdv * (centered_y + 1. / (num_classes - 1))
    cdf_plus = torch.sigmoid(plus_in)
    min_in = inv_stdv * (centered_y - 1. / (num_classes - 1))
    cdf_min = torch.sigmoid(min_in)

    # log probability for edge case of 0 (before scaling)
    # equivalent: torch.log(torch.sigmoid(plus_in))
    log_cdf_plus = plus_in - F.softplus(plus_in)

    # log probability for edge case of 255 (before scaling)
    # equivalent: (1 - torch.sigmoid(min_in)).log()

View source record →