Worked on enhancing the deployment workflow for diffusion-based models in the alibaba/MNN repository by improving the ONNX export path within the Diffusion Pipeline. Focused on simplifying the export process, the work involved removing dynamic axes and streamlining input parameters, which reduced export times and improved deployment reliability across environments. Addressed a specific issue with dynamic axes handling, leading to fewer export-time failures and contributing to a more robust deployment pipeline. Utilized Python scripting and model export techniques, with an emphasis on machine learning deployment. These targeted changes helped accelerate time-to-market and increased the stability of ONNX-based model exports.