HA WQ-V2: Hessian Aware trace-Weighted Quantization of Neural Networks