Developed a feature for the pytorch/pytorch repository to enhance multiprocessing signal handling under SLURM environments. The work introduced support for SIGUSR1 and SIGUSR2 signals, allowing configuration through environment variables to improve process lifecycle management in distributed high-performance computing workloads. Using Python, the implementation focused on robust signal handling and environment configuration, ensuring that worker processes could be managed more reliably. Comprehensive unit tests were added to validate signal behavior across multiple SLURM scenarios, addressing common issues with process termination and restart. This contribution improved the reliability and manageability of PyTorch’s multiprocessing module in HPC settings without fixing existing bugs.