In this paper we will present the result of such an endeavor in which a hybrid metrology solution was provided to measure individual CDs in GAA devices, using conventional scatterometry-based optical critical dimension (OCD) measurement combined with an advanced machine learning algorithm. Such an approach has been proven to provide a robust metrology solution for critical CD measurements in the entire Inner Spacer process module of GAA device. The current paper will discuss results from some of the Inner Spacer process modules such as Source/Drain Si Etch, SiGe Cavity Etch, Inner Spacer Etch, and Source/Drain Fine Etch. This hybrid model, called model-guided machine learning (MGML), uses rigorous coupled-wave analysis (RCWA) model to guide the machine learning engine. The superiority of the MGML model as compared to pure machine learning (ML) solution, in terms of robustness and correlation to blind test, is provided in this study.