Worked on stabilizing Falcon-180b text generation testing within the HabanaAI/optimum-habana-fork repository, focusing on improving continuous integration reliability. Addressed a critical reset error encountered during text generation tests by introducing a workaround flag, PT_HPU_DISABLE_ASYNC_COLLECTIVE, and refining the quantization configuration to ensure stable test execution. Utilized Python for model configuration and testing, applying targeted changes that enhanced the consistency and speed of CI test runs across the repository. The work centered on debugging and resolving a complex issue affecting automated testing, resulting in more reliable development workflows and improved maintainability for large-scale model evaluation in the project.