QuantEase, a layer-wise quantization framework, frames the problem as discrete-structured non-convex optimization. Our work leverages Coordinate Descent techniques, offering high-quality solutions without the need for matrix inversion or decomposition. - View it on GitHub