Architectural Pattern that is used to build a service: web-queue-worker
- Python >= 3.10
- Docker
- Tranformers
All the main dependencies and requirements are listed in requirements.txt file.
You can simply install all the dependencies via pip:
python3 -m pip install -r requirements.txt
It is recommended to have at least 16GB of RAM and dedicated GPU with at least 8GB VRAM.
For running the service you will also need Redis server on your local machine. The easiest way to run Redis is to use Docker container.
Open terminal and run:
$ docker run -d --name worker-redis -p 6379:6379 redis
After you will finish with exploration do not forget to stop the container
$ docker stop worker-redis
It is too resource consumable to run SD model directly via REST API server. The operation will take several minutes to consume a massive amount of memory and resources. To solve this, it is better to define another process, apart from server process, that will take care of this image generation task: so called worker. Its role to compute task in the background.
Worker is a process running continuously in the background, waiting for incoming tasks. The tasks are sent by the server. Also we need a communication channel between the server and the worker. This role will take up the queue. It will accept and stack messages coming from the server and will make them available to read for the worker.
Schema:
Prompt:
The Milky Way picture as it was taken from the space.
Output:
