Zhangfan contributed to the modelscope/ms-swift repository by enhancing documentation for training the Qwen3.5 model with Megatron-SWIFT. Focusing on AI model training workflows, Zhangfan authored best-practices guidance that clarified both full parameter and MTP training processes. The work emphasized clear Markdown documentation, aligning with repository standards to improve onboarding and reproducibility for users. By detailing setup steps and workflow nuances, Zhangfan reduced ambiguity for practitioners adopting Qwen3.5 training. Although no bugs were fixed during this period, the contribution demonstrated depth in technical writing and knowledge transfer, supporting the community in implementing robust AI model training practices with Megatron-SWIFT.