Worked on the PaddlePaddle/FastDeploy repository to address a documentation bug related to WINT2 quantization resource requirements for baidu/ERNIE-4.5-VL-424B-Paddle deployments. Focused on ensuring technical accuracy, the developer updated the supported_models.md file and its Chinese translation to reflect the correct minimum GPU VRAM, revising the value from 96G to 141G. This work involved careful review and editing of Markdown documentation, with attention to both English and Chinese language consistency. The contribution improved clarity for users deploying large models with WINT2 quantization, demonstrating a detail-oriented approach to documentation and cross-language technical communication within the project.