AI/ML is sending shockwaves through tech.
In the last few months, we've seen GitHub CoPilot, ChatGPT, and AI-generated content explode.
The core of this AI/ML magic is the model.
Determined AI makes building models faster and easier for ML engineers, data scientists, and researchers.
We want you to build and train an ML model with Determined.
Whether you are an experienced ML engineer or brand new to ML, Determined AI can help you get started.
Determined AI is an open source platform for building and training ML models that features:
- An easy-to-use web UI, a CLI tool, and multiple APIs
- Support for PyTorch, Tensorflow, Keras, and DeepSpeed
- Scalable hyperparameter search and visualization tools
- State-of-the-art optimization algorithms
- Reproduce experiments with artifact tracking
- Deploy your model with a built-in model registry
Determined AI can run on your machine, in the cloud, or on a supercomputer.
How do you get started with Determined AI?
- Check out our blog: Intro to Determined: A First Time User's Guide
- See our documentation: Quickstart for Model Developers
- Join our Slack community: Determined AI Community Slack
- Visit our GitHub: Determined AI GitHub
- See our YouTube Channel: Determined AI YouTube Channel
- Sign up for open office hours and lunch & learns: Determined AI Virtual Meetup Group
Determined AI runs best on devices with a performant CPU and/or NVIDIA CUDA-enabled GPU. We want to make this hackathon accessible to anyone, regardless of access to technology. Hackathon participants can request access to Determined Cloud, a managed hosted instance of Determined, with free access to GPU compute.
What to Build
A machine learning project (complete with deep learning-based model, training loop, evaluation loop, dataset) that works on the Determined AI platform. See the Rules.
What to Submit
- A model training project leveraging Determined AI
- A link to the public open-source code repository for the project, including:
- All code for the project
- Final model weights
- A README file containing:
- Project objective
- A data sample from your dataset with an explanation
- A description of your model architecture
- Instructions for how to run your training job
- A screenshot of your best metrics from the Determined web UI
- A description of your evaluation metrics
- Your evaluation results given these metrics, and
- Instructions for how to reproduce the evaluation results
- A link to each dataset used within the project. Datasets must be publicly available for the judges and sponsor to test your project.
- OPTIONAL: Include a demonstration video of your Project. The video portion of the submission:
- Should be around three (3) minutes
- Should include footage that shows the project functioning on the device for which it was built
- Must be uploaded to and made publicly visible on YouTube, Vimeo, Facebook Video, or Youku
- Must not include third-party trademarks, or copyrighted music or other material unless you have permission to use such material
$4,500 in prizes
$2,500 USD in cash
$1,000 USD in cash
Best Use of Distributed Training
$500 USD in cash
$500 USD in cash
Submitting to this hackathon could earn you:
Senior Director Of Engineering, Determined AI
Senior Developer Relations Manager for AI/ML, NVIDIA
Chief Strategy Officer, Infinite Retina
Chief Scientist, Pachyderm
Does the interaction with Determined AI and the API used demonstrate quality model development?
How big of an impact could the project have on the machine learning community?
How creative and unique is the project? Is the machine learning use case not commonly seen?
Use of Distributed Training
Does the Project show the best use of distributed training with Determined?