A browser interface for OneFlow Diffusers inspired by Stable Diffusion Web UI,the functions of txt2img and img2img are realized
docker run -it --rm \
--gpus all --ipc=host -p 7860:7860 --ulimit memlock=-1 --ulimit stack=67108864 \
-v ${PWD}:/workspace \
-v ${PWD}/.cache:/root/.cache \
-w /workspace \
oneflowinc/ai-painter:cu112 \
python3 launch.py --ip 0.0.0.0 --port 7860
Install depencies:
sh setup_env.sh
Launch the server:
python3 launch.py
If this fails, try the following two commands to install the latest oneflow
pip uninstall oneflow
python3 -m pip install -f https://staging.oneflow.info/branch/master/cu112 --pre oneflow
Then rerun the script.
There are other options besides ip
and port
mentioned above.
--ui-debug-mode
: launch without loading model--graph-mode
: use OneFlow graph mode which will accelerate the inference (but limited by fixed tensor shape)--device
: Target a specific device, eg:cuda:0
means the first GPU andcuda:1
means the second GPU