Linux

Please follow the primary README.md of this repo.

Windows

Windows users may stumble when installing the package triton. You can choose to run on CPU without xformers and triton installed.

To use CUDA, please refer to issue#24 to try solve the problem of triton installation.

MacOS

You can try to set up according to the following steps to use CPU or MPS device.

Install torch (Preview/Nighly version).

# MPS acceleration is available on MacOS 12.3+
pip install --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cpu

Check more details in official document.

Package triton and xformers is not needed since they work with CUDA. Remove the related packages.

Your requirements.txt should look like:

# requirements.txt
pytorch_lightning==1.4.2
einops
open-clip-torch
omegaconf
torchmetrics==0.6.0
opencv-python-headless
scipy
matplotlib
lpips
gradio
chardet
transformers
facexlib

pip install -r requirements.txt

Run the inference script and specify --device cpu or --device mps. Using MPS can accelarate your inference.

You can specify --tiled and related arguments to avoid OOM.