Libo Hao expanded PyTorch’s hardware support by porting distributed test cases to Intel GPU within the pytorch/pytorch repository, enabling compatibility and performance tracking for Intel devices. Using Python and leveraging expertise in distributed systems and GPU programming, Libo updated memory tracking to utilize a new accelerator interface, ensuring accurate cross-device memory statistics across CPU, CUDA, and Intel GPU. This work improved test reliability and observability, providing clearer insights for optimization and debugging. The integration of enhanced memory management and unit testing broadened PyTorch’s hardware coverage, addressing the need for robust validation and performance analysis on diverse accelerator platforms.