- experimental testnet for decentralized p2p AI processing
Setting up the environment:
Nvidia Driver:
sudo apt-get update && sudo apt-get install nvidia-driver-530
Docker:
- Linux
- Windows
- OSX
sudo docker run -p 31330:31330 --ipc host --gpus all --volume petals-cache:/cache --rm learningathome/petals:main python -m petals.cli.run_server bigscience/bloomz-petals --initial_peers /ip4/54.164.52.79/tcp/8989/p2p/QmR7G6GZzeQeJoXfTovp8PJdJ7qxKYFwXUa2fp9p6BYdDi --port 31330
Or install manually:
Anaconda Install:
wget https://repo.anaconda.com/archive/Anaconda3-2023.03-1-Linux-x86_64.sh
chmod +x ./Anaconda3-2023.03-1-Linux-x86_64.sh
./Anaconda3-2023.03-1-Linux-x86_64.sh
Answer yes to base env setup.
conda create -n pepeaitestnet python=3.10
conda activate pepeaitestnet
conda install pytorch pytorch-cuda=11.7 -c pytorch -c nvidia
pip install -U petals
Running a Node:
PepeAI net seperates processing by PepeCoin Nodes. Each node requires at least 1 GPU w/ 8GB+ or more of memory.
- bloom model layers distrubuted across fleet of pepecoin nodes
- will support various models as they are updated for pepecoin ai net
Start GPU Processing Node:
- ensure environment is setup as indicated above
- If you have more than 1 GPU, set the GPU index before starting the node
- you can run multiple nodes under screen
- optional IP:PORT for improved network performance, use --public_ip x.x.x.x --port 8989
- from linux cli, to get public ip
curl ifconfig.me
- Run Node (don't forget to add --port 8989 --public_ip x.x.x.x) if you have setup port forwarding on your router
export CUDA_VISIBLE_DEVICES=0 # Insert the GPU index here, counting from zero
python -m petals.cli.run_server bigscience/bloomz-petals --initial_peers /ip4/54.164.52.79/tcp/8989/p2p/QmR7G6GZzeQeJoXfTovp8PJdJ7qxKYFwXUa2fp9p6BYdDi
Demo Chatbot:
python3 ./pepeaichatbot.py