- Docker for training condition diffusion model
- 2024/09: Internal testing starting.
- Put the images into
traindata
folder (both png and jpeg formats are fine).
├──traindata
├──1.jpg
├──2.jpg
├──3.jpg
- Generate a
metadata.csv
file including the metadata.
Image | Ethnicity | Sex | Age | DR |
---|---|---|---|---|
1.jpg | British | female | 62 | mild diabetic retinopathy |
2.jpg | Chinese | female | 71 | moderate diabetic retinopathy |
3.jpg | Indian | male | 52 | no diabetic retinopathy |
If you have not installed Docker on your machine. Please follow official instructions.
- Download the docker
docker pull yukunzhou/diffuser_docker
- Run it
Please substitute the {Absolute_path}
with the path to traindata
folder.
docker run --gpus all -it -v {Absolute_path}:/app/diffusers/Global_DM yukunzhou/diffuser_docker
e.g.docker run --gpus all -it -v /home/yukun:/app/diffusers/Global_DM yukunzhou/diffuser_docker
The process will generate a diffusion1.5
,diffusion2
, and sdxl
folder in {Absolute_path}
path. The model weights will be saved there.
TBC