Get decoder log likelihood questions

[Ahmad-Omar-Ahsan]

Hello all, I have a question regarding the _get_decoder_log_likelihood function of the DiffusionInferer

    def _get_decoder_log_likelihood(
        self,
        inputs: torch.Tensor,
        means: torch.Tensor,
        log_scales: torch.Tensor,
        original_input_range: tuple = (0, 255),
        scaled_input_range: tuple = (0, 1),
    ) -> torch.Tensor:
        """
        Compute the log-likelihood of a Gaussian distribution discretizing to a
        given image. Code adapted from https://github.com/openai/improved-diffusion.

        Args:
            input: the target images. It is assumed that this was uint8 values,
                      rescaled to the range [-1, 1].
            means: the Gaussian mean Tensor.
            log_scales: the Gaussian log stddev Tensor.
            original_input_range: the [min,max] intensity range of the input data before any scaling was applied.
            scaled_input_range: the [min,max] intensity range of the input data after scaling.
        """
        if inputs.shape != means.shape:
            raise ValueError(f"Inputs and means must have the same shape, got {inputs.shape} and {means.shape}")
        bin_width = (scaled_input_range[1] - scaled_input_range[0]) / (
            original_input_range[1] - original_input_range[0]
        )
        centered_x = inputs - means
        inv_stdv = torch.exp(-log_scales)
        plus_in = inv_stdv * (centered_x + bin_width / 2)
        cdf_plus = self._approx_standard_normal_cdf(plus_in)
        min_in = inv_stdv * (centered_x - bin_width / 2)
        cdf_min = self._approx_standard_normal_cdf(min_in)
        log_cdf_plus = torch.log(cdf_plus.clamp(min=1e-12))
        log_one_minus_cdf_min = torch.log((1.0 - cdf_min).clamp(min=1e-12))
        cdf_delta = cdf_plus - cdf_min
        log_probs = torch.where(
            inputs < -0.999,
            log_cdf_plus,
            torch.where(inputs > 0.999, log_one_minus_cdf_min, torch.log(cdf_delta.clamp(min=1e-12))),
        )
        return log_probs

The documentation says that it expects the input to be between -1 to 1 after scaling. However, in the arguments, it expects the scaled input_range to be between 0,1. Does it really matter if it is between 0 to 1 or -1 to 1?

Here is the link to the code.