Worked on the deepspeedai/DeepSpeed repository, delivering a targeted bug fix to the Autotuning module using Python. Addressed a typo in the was_interrupted function path and corrected the memory unit divisor from 1014 to 1024, ensuring accurate memory calculations during autotuning runs. This patch improved the reliability of performance measurements and memory budgeting, directly impacting the stability of optimization workflows. The work also included code refactoring to align function naming with actual behavior, enhancing code readability and maintainability. Focused on bug fixing and code clarity, the contribution strengthened the correctness and dependability of the autotuning process for users.