Jingya Huang
Machine Learning Engineer at Hugging Face
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolJingya Huang is a Machine Learning Engineer at Hugging Face with five years of experience bridging research-grade models and production-efficient deployments. Trained originally in nuclear science and refined with MS degrees in Machine Learning and Computer Science from Sorbonne and PSL/MINES ParisTech, she focuses on making high-performing ML models resource-efficient and accessible. At Hugging Face she has driven ONNX Runtime integration and trainer implementations in flagship repos like optimum and transformers, improving inference/training speed and CUDA/FP16 compatibility for models such as GPT-2 and DeBERTa. Her background in nuclear engineering and short stint as an operator give her a pragmatic, systems-oriented approach to optimization that complements her open-source impact. Based in Paris, she actively documents work on GitHub and prioritizes democratizing ML without “devouring” compute resources.
5 years of coding experience
1 year of employment as a software developer
Bachelor Degree in nuclear science and technology, Nuclear science and technology, Bachelor Degree in nuclear science and technology, Nuclear science and technology at 中山大学It’s a high school, It’s a high school at Shenzhen Middle SchoolMaster of Science - MS, Computer and Information Sciences, General, Master of Science - MS, Computer and Information Sciences, General at PSL Research UniversityMaster of Science - MS, Machine Learning, Master of Science - MS, Machine Learning at Sorbonne UniversityMaster of Science - MS / Diplôme d'ingénieur, Master of Science - MS / Diplôme d'ingénieur at MINES ParisTech
English, French, Chinese
Github Skills (26)
transformers10
pytorch10
python10
machine-learning10
inference10
onnx10
deep-learning10
onnxruntime10
optimisation10
nlp10
optimization10
machine-learning-models9
ml-deployment9
gpt9
cuda9
Programming languages (6)
Github contributions (5)
huggingface/optimum
Mar 2022 - Jan 2023
🚀 Accelerate inference and training of 🤗 Transformers, Diffusers, TIMM and Sentence Transformers with easy to use hardware optimization tools
Role in this project:
ML Engineer
ML EngineerContributions:3 releases, 386 reviews, 83 commits in 10 months
Contributions summary:Jingya's primary contribution involves developing and integrating ONNX Runtime (ORT) for accelerating training and inference within the Hugging Face optimum library. They added ORTTrainer, Seq2SeqORTTrainer, and other trainer variants. Their work included creating examples and tests, focusing on question answering, translation, and language modeling tasks. The user also implemented IOBinding support, improved compatibility, and optimized the library to support diverse models.
trainingoptimization-toolstransformersintelhardware
huggingface/transformers
May 2022 - Dec 2022
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:30 reviews, 5 commits, 11 PRs in 6 months
Contributions summary:Jingya's contributions primarily focus on enhancing the ONNX export functionality and addressing issues related to model training and deployment, especially concerning the integration with CUDA devices. This includes adding CUDA support for ONNX export and resolving compatibility issues with different data types in the context of FP16 training, specifically for GPT2 and DeBERTa models. Furthermore, the user has addressed issues when tracing is enabled in GPT2. Additional changes involve supporting subfolders within AutoProcessor and addressing matmul inputs dtype issues, demonstrating a commitment to model compatibility and performance optimization.
pythonbertspeech-recognitionstate-of-the-artflax
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.