This is a template for running a FLock Training Node on Akash. It enables users to run FLock's Training Node Quickstart Script directly from the Akash Console and earn FML rewards on train.flock.io.
train.flock.io is the gateway to FLock.io's decentralized AI training platform, AI Arena. It is currently on incentivised testnet, and all participants who have earned FML rewards will receive mainnet airdrops.
To participate, you need to first get whitelisted, acquire FML test tokens and test tokens for Base Sepolia, then stake FML on the task you wish to train models for. Afterwards, you can use this template to run training tasks with Akash compute; the script automates the entire Training Node process, from downloading training dataset, model training, uploading to a Hugging Face repo, and submitting the training task.
🚀 About FLock.io
FLock.io is a decentralised AI co-creation platform aimed at democratising AI development and alignment, battling the concentration of power and data ownership in centralised corporations. It closed a $6m seed round in March 2024, led by Lightspeed Faction with participation from DCG, Volt, OKX Ventures among others. The core team, from Oxford University, has 10+ years of AI and blockchain experience.
To use the training node,obtain your FLOCK_API_KEY and task_id from FLock,obtain your HG_USERNAME and HF_TOKEN from Hugging Face and fill in the deploy.yaml file accordingly.To train with your own dataset or configure your training parameters, upload demo_data.jsonl and training_args.yml to your GitHub project (see example), and then enter the project URL in the GIT_URL field in deploy.yaml.
For comprehensive guides and documentation, refer to the FLock Docs.
If you have any questions, encounter any issues, or have feature requests, feel free to reach out:
- Follow us on Twitter
- Connect with us on LinkedIn
- Join the discussion on Discord
- Join our Telegram community