Function dequantize

pub fn dequantize<R, F>(tensor: CubeTensor<R>) -> CubeTensor<R>
where
    R: CubeRuntime,
    F: CubeElement,

Convert the tensor back to a higher precision data type.