Creating a Tensor with a value that is already encoded in bfloat16

Hi,
I am very new to PyTorch. But I would like to use its features to convert between datatypes.

Suppose I have a hex number that is already encoded in bfloat16 = 0xABCD

torch.tensor([0xABCD], dType=bfloat16) → this seems to treat 0xABCD as a hex encoded version of the decimal number 0d7117.

How can I initialise a variable that is already encoded in bfloat16 format and then have PyTorch convert it to its decimal value?