Skip to content

Latest commit

 

History

History
80 lines (64 loc) · 2.29 KB

README.md

File metadata and controls

80 lines (64 loc) · 2.29 KB

Deploy Stable Diffusion on Google Cloud

We follow this guide: AUTOMATIC1111/stable-diffusion-webui#9131

Note that in this example, we just deploy with cpu.

0. Prerequisites

  • Set up a Google Cloud billing account. It neccessery for you to create a VM with GPU (If you don't want to use GPU, you are ok with free trial).
  • Setup a firewall rule for private access
    • Targets: Specified target tags; Target tags: diffusion-demo (You need to add this tag to VM instance).
    • Source filter: IPv4 ranges -> Put your IP range in Source range box (You can use 0.0.0.0/0 to open for everyone in the Internet).
    • Protocols and ports: TCP -> Put your port you want to open in Ports box (e.g 8080)

1. Create VM Instance

  • Create a VM instance based on your demand. Example:

    • 1 GPU NVIDIA T4
    • n1-standard-4 (4 vCPU, 15 GB memory)
    • Ubuntu 22.04 LTS
    • 128GB Storage

  • In the Firewall section: Allow HTTP traffic

  • Optional: Add your SSH key to instance:

    • Create your ssh-key in local machine with: ssh-keygen
    • Copy public key in your_fine_name.pub file and add to your instance: Click to instance -> Edit -> SSH Keys -> Add Item -> Paste your public key -> Save
    • Access your VM with SSH.

2. Deploy on VM Instance

2.1 Setup

  • Python
sudo apt-get update
sudo apt install python3-pip
  • GPU driver:
sudo apt install ubuntu-drivers-common
sudo ubuntu-drivers autoinstall

# Reboot
sudo reboot
  • (Optional) Miniconda
curl -sL "https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh" > "Miniconda3.sh"

bash Miniconda3.sh
  • Docker
sudo snap install docker
sudo snap start docker

2.2 Run with Docker

Install

git clone https://github.com/mles-02/diffustion-optimize.git
cd diffustion-optimize/deployment/fastapi

Create a .env file (we provided example .env.example). Change HOST and PORT in Dockerfile.cpu and docker-compose.yaml file (note that port is the port that you exposed in fireware rule). Then run docker-compose:

sudo docker-compose -d --build

View log with

sudo docker-compose logs -f