Rotation Invariant Quantization for Model Compression