Worked on enhancing the Lightning-AI/torchmetrics repository by improving documentation for the VisualInformationFidelity (VIF) metric. Focused on clarifying the required input shape for VIF, the update specified that both predicted and ground truth images must be at least 41 by 41 pixels, and detailed the ValueError raised for non-conforming inputs. This documentation update, implemented in Python, aligned the docstring with the actual validation logic, reducing confusion for developers integrating VIF into their workflows. By emphasizing clear API usage and error behavior, the work aimed to streamline onboarding and decrease support requests related to input handling for this metric.