don't disagree, but there is a big difference between 'quantized model/weights' and quantized activations