Has anyone tried this quantization method on Regression mode

Has anyone tried this quantization method on Regression mode? If so how much performance was degraded with the quantized model?