StyleGAN FastAPI

Simplified Web REST API of the StyleGAN using FastAPI

This is a web backend application to make the generator models trained by Precure StyleGAN ADA available on the web in a simplified manner. An example implementation of the front-end client is StyleGAN Vue and is deployed here for immediate trial.

Requirements

To run this application, you need a Unix-like system with Python >= 3.12 installed. Precure StyleGAN ADA's dependencies.

Use the requirements.txt to install minimal dependencies for serving and inferencing.

pip3 install -r requirements.txt

While Uvicorn is recommended as ASGI middleware, other alternatives such as Hypercorn also work.

In a product environment, a typical example is process replication using a combination of Gunicorn and Uvicorn.

Quick Start: Run the Default Server

This repository includes the default server configuration using Uvicorn with some trained models. You can try out quickly as below.

git clone --recursive https://github.com/curegit/stylegan-fastapi.git
cd stylegan-fastapi
pip3 install -r requirements.txt
uvicorn main:app --reload

Note that it is only for trial or debugging, and not suitable for production because it is not process-replication-enabled.

API

See the

Configuration

StyleGAN FastAPI can be customized by using a configuration file. StyleGAN FastAPI uses the TOML format for configuration, and the default TOML file is located in default/config.toml.

You can specify a different TOML file to use by setting the STYLEGAN_TOML environment variable. STYLEGAN_TOML must be an absolute path or a relative path from the working directory to the TOML file. Note that the STYLEGAN_TOML environment variable must be set before importing the api Python package.

export STYLEGAN_TOML="./custom_config.toml"
uvicorn main:app

Configuration TOML Specification

standard TOML complete raw schema definition in Python api/conf.py

All properties are optional except model files You must specify at least one model.

General Settings (root table)

Key	Type	Description
`title`	string	The title of the software.
`version`	string	The version number of the software.
`description`	string	A brief description of the software.
`lossy`	boolean	A boolean value indicating whether lossy compression is used. The default is `false`.
`docs`	boolean	A boolean value indicating whether documentation is enabled.
`redoc`	boolean	A boolean value indicating whether ReDoc is enabled.
`server`	ServerConfig	dd
`models`	{[key: string]: ModelConfig}	d

Server Settings

Key	Type	Description
`gpu`	boolean or integer	A boolean value indicating whether GPU is enabled.
`logger`	string	The name of the logger used.
`tmp_dir`	string	The path to the temporary directory.
`http`	HTTPConfig	The path to the temporary directory.
`limit`	LimitConfig	The path to the temporary directory.

HTTP Settings

HTTP Settings	Type	Description
`forwarded`	boolean	A boolean value indicating whether HTTP forwarding is enabled.
`forwarded_headers`	string[]
`cors`	CORSConfig	A boolean value indicating whether HTTP forwarding is enabled.

CORS Settings

`CORSConfig`	Type	Description
`enabled`	boolean	A boolean value indicating whether CORS is enabled.
`origins`	string[]	A list of allowed origins.

Limit Settings

Limit Settings	Type	Description
`min_delay`	float	The minimum delay time.
`block`	SignallingBlockConfig	A block of settings for blocking requests.
`concurrency`	ConcurrencyLimitConfig	A block of settings for concurrency.
`rate`	RateLimitConfig	A block of settings for rate limiting.

SignallingBlockConfig

Limit Settings	Type	Description
`enabled`	boolean	A block of settings for rate limiting.
`timeout`	float	A block of settings for rate limiting.
`poll`	float	A block of settings for rate limiting.

ConcurrencyLimitConfig

Limit Settings	Type	Description
`enabled`	boolean	A block of settings for rate limiting.
`max_concurrency`	int	A block of settings for rate limiting.
`max_queue`	int	A block of settings for rate limiting.
`timeout`	float	A block of settings for rate limiting.
`poll`	float	A block of settings for rate limiting.

RateLimitConfig

Limit Settings	Type	Description
`enabled`	boolean	A block of settings for rate limiting.
`window`	float	A block of settings for rate limiting.
`max_request`	int	A block of settings for rate limiting.

Model Settings

Model Settings	Type	Description
`file`	string	models: relative path from working directory. The path to the model file.
`relative`	boolean	A boolean value indicating whether the path is relative.
`name`	string	The name of the model.
`description`	string	A brief description of the model.
`lossy`	boolean	A boolean value indicating whether lossy compression is used.
`gpu`	boolean or int	A boolean value indicating whether GPU is enabled.

Gunicorn

Example

`stylegan.service`

[Unit]
Description=Gunicorn StyleGAN Daemon
Requires=stylegan.socket
After=network.target

[Service]
Type=notify
User=spam
Group=spam
WorkingDirectory=/home/spam/stylegan
ExecStart=/usr/local/bin/python3.12 -B -m gunicorn -c gunicorn.config.py
RuntimeDirectory=stylegan
PrivateTmp=true
ExecReload=/bin/kill -s HUP $MAINPID
KillMode=mixed
TimeoutStopSec=30
Restart=always

[Install]
WantedBy=multi-user.target

`stylegan.socket`

[Unit]
Description=Gunicorn StyleGAN Socket

[Socket]
ListenStream=/run/stylegan.sock
SocketUser=www-data
SocketMode=600

[Install]
WantedBy=sockets.target

Notes

set proxy header correctly For production (reverse proxy layer) It is also recommeded to set limit header and body size

License

CC BY-NC 4.0