carefree-creator is built on top of carefree-learn, and requires:

• Python>=3.8
• pytorch>=1.12.0. Please refer to PyTorch's official website, and it is highly recommended to pre-install PyTorch with conda.

Related issue: #10.

This project will eat up 11~13 GB of GPU RAM if no modifications are made, because it actually integrates FIVE different SD versions together, and many other models as well. 🤣

There are two ways that can reduce the usage of GPU RAM - lazy loading and partial loading, see the following Run section for more details.

pip install carefree-creator

If you are interested in the latest features, you may use pip to install from source as well:

git clone https://github.com/carefree0910/carefree-creator.git
cd carefree-creator
pip install -e .

carefree-creator builds a CLI for you to setup your local service. For instance, we can:

cfcreator serve

If you don't have an NVIDIA GPU (e.g. mac), you may try:

cfcreator serve --cpu

If you are using your GPU-powered laptop, you may try:

cfcreator serve --limit 1

The --limit flag is used to limit the number of loading models. By specifying 1, only the executing model will be loaded, and other models will stay on your disk.

See #10 for more details.

If you have plenty of RAM resources but your GPU RAM is not large enough, you may try:

cfcreator serve --lazy

With the --lazy flag, the models will be loaded to RAM, and only the executing model will be moved to GPU RAM.

So as an exchange, your RAM will be eaten up! 🤣

If you only want to try the SD basic endpoints, you may use:

cfcreator serve --focus sd.base

And if you only want to try the SD anime endpoints, you may use:

cfcreator serve --focus sd.anime

More usages could be found by:

cfcreator serve --help

export TAG_NAME=cfcreator
git clone https://github.com/carefree0910/carefree-creator.git
cd carefree-creator

docker build -t $TAG_NAME .

If your internet environment lands in China, it might be faster to build with Dockerfile.cn:

docker build -t $TAG_NAME -f Dockerfile.cn .

docker run --gpus all --rm -p 8123:8123 $TAG_NAME:latest

• Stable Diffusion, the foundation of various generation methods.
• Stable Diffusion from runwayml, the adopted SD-inpainting method.
• Waifu Diffusion, the anime-finetuned version of Stable Diffusion.
• Real ESRGAN, the adopted Super Resolution methods.
• Latent Diffusion, the adopted Inpainting & Landscape Synthesis method.
• carefree-learn, the code base that has re-implemented all the models above and provided clean and handy APIs.
• And You! Thank you for watching!