Function dequantize 

pub fn dequantize(
    tensor: &Tensor<i8>,
    params: &QuantizationParams,
) -> Result<Tensor<f32>>

Dequantize an INT8 or INT4 tensor back to float32