Worked on the vllm-project/llm-compressor repository to enhance the reliability of model compression workflows by expanding unit test coverage for the ModuleSparsificationInfo class. Focused on validating the params_quantized_percent and params_sparse_percent methods, the work aimed to improve the accuracy of quantization and sparsity parameter calculations and reduce the risk of regressions. Leveraged Python, PyTorch, and Pytest to implement a comprehensive suite of unit tests, ensuring that sparsification features are robustly validated. This approach supports safer deployments and streamlines future debugging efforts by catching potential issues early, contributing to higher code quality and more maintainable model compression pipelines.