vast.ai nomic@ubuntu provisioning script

provision.sh

# generate an api key so others don't abuse this server; use it to authenticate the API calls
export MODEL_KEY="$(openssl rand -base64 24)"

# recommended by the docs
env >> /etc/environment

# update the OS
export DEBIAN_FRONTEND=noninteractive
apt update; apt upgrade -y
apt install -y nvtop

# create a server runner script  
echo '#!/bin/bash' >~/run-server
echo "echo === Your API key is: $MODEL_KEY ===" >>~/run-server
echo docker run -d --restart unless-stopped --gpus all --name llama-server -v ~/models:/models -p 8000:8000 ghcr.io/ggml-org/llama.cpp:server-cuda -m "/models/$MODEL_NAME" $MODEL_FLAGS --port 8000 --host 0.0.0.0 --n-gpu-layers 999 --api-key "$MODEL_KEY" >>~/run-server
chmod 700 ~/run-server

# install FAR manager for convenient administration 
wget https://github.com/spvkgn/far2l-portable/releases/download/latest/far2l-x86_64.run.tar; tar -xf far2l-x86_64.run.tar; cp "$(tar -tf far2l-x86_64.run.tar)" /usr/sbin/far2l; chmod 755 /usr/sbin/far2l; echo "far2l; exit" >>~/.bash_profile

# download the model and the container
mkdir ~/models; wget -O "$HOME/models/$MODEL_NAME" "$MODEL_URL"
docker pull ghcr.io/ggml-org/llama.cpp:server-cuda

# run the server
~/run-server