Add support for passing pytorch tensors directly to kernel calls

[NaderAlAwar]

This checks if torch is installed and uses torch.is_tensor() to check if the passed argument is a pytorch tensor.