I'm running into an issue with an ablation study for a paper I'm preparing. I trained a model. The model achieved my best result, and I saved the trained checkpoint (.pth file). Now my supervisor wants me to perform an ablation study by removing components and how it impacts the accuracy. My concern
