Dive into the AI SpeedRacers Quest, your fast track to mastering ML model optimization on the Intel Developer Cloud. This challenge is your gateway to applying Intel’s advanced techniques for peak efficiency and performance for AI challenges, that could be focused on:
- LLM Fine-Tuning Pipeline Optimization
- Optimized classical Machine Learning and Deep Learning
- Launching a RAG Chatbot
- Applying an quantization to your model
- And we're not limiting your ideas, just pitch and show us your great idea!
Achievement Unlocked: Craft an optimized model on Intel Developer Cloud, then spotlight it on the HuggingFace model hub, complete with a model card detailing its journey and capabilities on the Intel Developer Cloud.
Your HuggingFace model card will feature:
- Model Essentials: Name, description, license, and purpose.
- How-To: Usage instructions.
- Know-How: Limitations and optimizations.
- Proof of Excellence: Key metrics like Accuracy and MAE.
Embark on the AI SpeedRacers Quest to not only refine your skills but also to leave a mark on the AI community with your optimized, efficient ML model.
- Intel Developer Cloud guide
- PredictionGuard LLM/RAG Guide
- Learn more about Intel AI optimizations for Pandas, Scikit-learn, Pytorch, Tensorflow here
- Learn here about PredictionGuard API if you want to work with LLM/RAG API along with important safeguard (like factual consistency checking, toxicity filters, PII filters, prompt injection detection, etc.)
-
Development on Intel Developer Cloud (25%)
-
0 pts - No use of Intel® Developer Cloud; the solution is not valid for judging.
-
10 pts – Use of IDC Jupyter Notebook for development
-
20 pts – Complete development and showcase on IDC aligning with hackathon challenge.
-
-
Open Source (25%)
-
0 pts – The project is not open source.
-
10 pts – Project is opensource on GitHub/external repo; IP-based exceptions allowed.
-
20 pts – Projects are open source and pushed to Hugging Face.
-
-
Problem Solving (25%)
-
0 pts – The project does not show any creative solution or solving of a real-world use case.
-
10 pts – The project shows at minimum some creative use of AI and UI elements to solve a real-world use case.
-
20 pts – The project has excellent UI and clearly solves a real-world use case. Uses AI and take aspects like LLM hallucinations, bias, etc into consideration.
-
-
Optimization (25%)
-
0 pts – The project shows no use of Intel optimized libraries and optimizations.
-
10 pts – Some use of Intel optimized libraries and optimizations.
-
20 pts – Extensive use of Intel optimized libraries and optimizations, additional optimization techniques applied.
-
🖲️ Intel Judges can award fractional points at will, based on their evaluation of the project. Intel judges retains complete discretion in allocating points for each category.
Vladimir Kilyazov will be glad to answer your questions during the Deep Dive. We’ll also be available on Discord.