Convert numpy array with image data to a tensor with method pre_process()

[TurboJonte]

I need to use the PyTorch inferencer to gain CUDA support and the PyTorch inferencer only accepts an tensor as input compared to the OpenVINO inferencer which takes a numpa array.
I thus tried to use the pre_process method of TorchInferencer to convert it but I get the error:

AttributeError: 'numpy.ndarray' object has no attribute 'to'

So this method can't really work as it is designed.

def pre_process(self, image: np.ndarray) -> torch.Tensor:
        """Pre process the input image.

        Args:
            image (np.ndarray): Input image

        Returns:
            Tensor: pre-processed image.
        """
        if len(image) == 3:
            image = image.unsqueeze(0)

        return image.to(self.device)

[TurboJonte]

I just found that this is addressed in #1816