Moto focused on improving the reliability of quantization-aware training workflows in the pytorch/pytorch repository by addressing a critical issue with pickle serialization for the _ExperimentalConfig class. Using Python and leveraging expertise in data serialization and unit testing, Moto corrected the __getstate__ and __setstate__ methods to ensure all serialized fields were properly handled and aligned with constructor parameters. This fix resolved deep-copy failures that previously blocked QAT workflows, particularly in downstream projects like d2go. The work enhanced the robustness and reproducibility of configuration state restoration, demonstrating careful attention to serialization semantics and thorough validation through targeted unit tests.