Hong Zhang
Professor Of Physics at Zhejiang University
China
Summary
🤩
RockstarHong Zhang is a Professor of Physics at Zhejiang University with seven years of cross-disciplinary experience bridging academic research and machine learning engineering. Though rooted in physics, he has become a prolific contributor to large-scale open-source ML tooling, notably expanding model support and training/inference workflows in the widely used modelscope/ms-swift project for hundreds of LLMs and multimodal models. He excels at integrating new architectures (e.g., Molmo, Emu3), improving dataset handling and sampling, and building practical scripts that move models from experimentation to deployment. Based in China, he combines rigorous scientific thinking with hands-on software development, enabling reproducible ML experiments at scale. Colleagues value him for translating complex theoretical problems into robust engineering solutions that support diverse model families.
7 years of coding experience
Github Skills (10)
transformers10
pytorch10
machine-learning10
lm10
inference10
trainings10
large-language-models10
python10
modeling10
llm10
Programming languages (2)
Github contributions (5)
modelscope/ms-swift
Oct 2024 - Mar 2025
Use PEFT or Full-parameter to finetune 500+ LLMs (Qwen2.5, InternLM3, GLM4, Llama3.3, Mistral, Yi1.5, Baichuan2, DeepSeek-R1, ...) and 200+ MLLMs (Qwen2.5-VL, Qwen2-Audio, Llama3.2-Vision, Llava, InternVL2.5, MiniCPM-V-2.6, GLM4v, Xcomposer2.5, Yi-VL, DeepSeek-VL2, Phi3.5-Vision, GOT-OCR2, ...).
Role in this project:
ML Engineer
ML EngineerContributions:7 reviews, 30 PRs, 12 pushes in 5 months
Contributions summary:Hong's commits primarily focused on expanding the supported models and functionalities within the `ms-swift` project, specifically related to large language models (LLMs) and multi-modal models. Their contributions involved adding support for new model types, including Molmo and Emu3, by integrating relevant templates and configurations. Furthermore, the user implemented scripts for training and inference, demonstrating a strong grasp of model integration and deployment within the swift framework. The commits also included improvements to dataset support and sampling techniques.
deepseek-r1deploydistillembeddinggrpo
mi804/ros_nav
Mar 2019 - May 2019
Contributions:14 commits, 14 pushes, 1 branch in 1 month
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.