milady themed operations runner with optimized defaults for maximizing cpu+ram+gpu utilization accross a cluster
DISCLAIMER: This tool encapsulates what could be considered "best practices" for... bad practices.
As a distributed computing framework, MiladyOS enables the remote execution of arbitrary code. You should only install MiladyOS workers within networks that you trust. This is standard among distributed computing frameworks, but is worth repeating.
The client is able to download runners, greet milady neibours, create decentralized overlay networks, which then performs any number of Inference/Training tasks. This can potentially be used in a bad manner. Run the client with the least priviliges where possble.
Find more information about the "principle of least privilege" on wikipedia: https://en.wikipedia.org/wiki/Principle_of_least_privilege
You should have Nvidia drivers installed and nvidia-container-toolkit configured, post which it's recommended to reboot once prior to running this.
curl -sSL https://raw.githubusercontent.com/theycallmeloki/MiladyOS/main/install_miladyos.sh | bash
The default username is admin and the default password is password
# docker buildx build --platform linux/amd64,linux/arm64 -t ogmiladyloki/miladyos --push .
Look for the Caddyfile in the repo and update it to point to your domain You would also set the A record to point to your IP address on your domain provider
yourfancydomain.com
TODO: Drop edith-cli dependency
Pachyderm bootrstrapped from edith-cli
You will need this script at the end of the edith-cli installation
Add the following to the Dockerfile in the root section and build yourself a container, checkout builder.sh
# Add kubeconfig file to the docker
ADD kubeconfig.yaml /root/.kube/config
# Add Pachctl config to the container
COPY config.json /root/.pachyderm/config.json