Worked on the pytorch/ao repository to enhance the Wanda Sparsifier by introducing configurable layer observers for model quantization. Developed a system in Python using PyTorch that allows observers to be attached to specific model layers based on an optional configuration, improving flexibility in quantization workflows. Addressed observer attachment logic to ensure correct behavior across different configuration scenarios, supporting both custom and no-config use cases. Added comprehensive unit tests to validate these enhancements, focusing on robustness and usability for model developers. The work deepened quantization tooling, enabling targeted sparsification and facilitating easier experimentation with layer-level observer configurations in deep learning models.