Foundry-rit-ai-training-model-challenge

🚀 Foundry-RIT AI Training Model Challenge: Run a FLock Training Node on Akash

Welcome to the Foundry-RIT AI Training Model Challenge! This repository serves as your template for running a FLock Training Node on Akash, as part of our exciting competition running from October 1 to November 12.

🌟 About the Challenge

The Foundry-RIT AI Training Model Challenge is a collaborative effort between Foundry Digital and RIT's AI Club. This competition aims to introduce students to cutting-edge AI technologies and decentralized computing platforms. Participants will use FLock.io's federated learning platform and Akash Network's decentralized compute resources provided by Foundry Digital.

🛠 What You'll Be Doing

As a participant, you'll:

1. Run a Flock Training Node on Akash
2. Train AI models using federated learning techniques
3. Compete to create the best-performing models
4. Earn FML rewards on train.flock.io

🔧 Getting Started

Prerequisites

1. Get whitelisted on train.flock.io
2. Acquire FML test tokens and Base Sepolia test tokens
3. Stake FML on your chosen training task

Setup

1. Fork this repository
2. Obtain your FLOCK_API_KEY and task_id from Flock
3. Get your HG_USERNAME and HF_TOKEN from Hugging Face
4. Fill in the deploy.yaml file with your credentials and chosen task

Customization (Optional)

To train with your own dataset or configure training parameters:

1. Upload your demo_data.jsonl and training_args.yml to your forked repository
2. Update the GIT_URL field in deploy.yaml with your repository URL

📚 Resources

• Flock Documentation
• Akash Network Documentation

📅 Important Dates

• Competition Start: October 1, 2024
• Submission Deadline: November 5, 2024
• Winners Announcement: November 12, 2024

🤝 Support

If you have questions or need assistance:

• Join our Discord channel
• Attend our weekly office hours (schedule to be announced)