Can it support docker

[jw-star]

Can it support docker

[AUTOMATIC1111]

maybe

[Pipazoul]

Here a a version i've made, probably not the most optimized, I will make a pull request soon

To build you need to have in the same folder :

• sd-v1-4.ckpt
• webui.py

FROM pytorch/pytorch:latest

# Make RUN commands use `bash --login`:
SHELL ["/bin/bash", "--login", "-c"]

RUN apt update
RUN apt install git curl -y

RUN apt-get update -yq \
    && apt-get -yq install curl gnupg ca-certificates \
    && curl -L https://deb.nodesource.com/setup_12.x | bash \
    && apt-get update -yq \
    && apt-get install -yq \
        nodejs

RUN apt install -y wget

RUN npm install -g localtunnel

WORKDIR /app

RUN git clone https://github.com/CompVis/stable-diffusion.git
WORKDIR /app/stable-diffusion

RUN conda env create -f environment.yaml
COPY webui.py ./
RUN mkdir models/ldm/stable-diffusion-v1/
RUN mkdir embeddings
RUN apt install wget -y
RUN wget -P embeddings/ url-to-my-emmbeding.pt


COPY sd-v1-4.ckpt /app/stable-diffusion/models/ldm/stable-diffusion-v1/model.ckpt


SHELL ["conda", "run", "-n", "ldm", "/bin/bash", "-c"]

RUN git clone https://github.com/TencentARC/GFPGAN.git
WORKDIR /app/stable-diffusion/GFPGAN
RUN pip install -r requirements.txt
RUN wget -P experiments/pretrained_models/ https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth



WORKDIR /app/stable-diffusion/
RUN pip install opencv-python
RUN apt update && apt install -y libsm6 libxext6
RUN apt-get install -y libxrender-dev
RUN pip install realesrgan

RUN pip install gradio
RUN pip install git+https://github.com/crowsonkb/k-diffusion/


RUN mkdir outputs
RUN mkdir outputs/txt2img-samples
RUN mkdir outputs/img2img-samples
RUN mkdir outputs/extras-samples
RUN mkdir outputs/txt2img-samples/samples
RUN mkdir outputs/img2img-samples/samples
RUN mkdir outputs/extras-samples/samples

in the docker image you can launch the following commands
I've used local tunnel, to expose my app because I rent a distant GPU

cd /app/stable-diffusion
nohup bash -c "lt --port 8080 &" && sleep 4
conda run -n ldm python webui.py --gfpgan-dir GFPGAN --no-progressbar-hiding 

[Pipazoul]

You also can use the gradio share url in replacement of localtunnel

[jw-star]

thanks

[AndreaPi]

@Pipazoul are you still planning to make a PR? This could be very useful. I would also suggest adding an entrypoint.sh file that executes the commands you should otherwise run in the Docker container. In other words, you could modify the Dockerfile adding these lines:

EXPOSE 8080
COPY ./entrypoint.sh /sd/
ENTRYPOINT /sd/entrypoint.sh

and include an entrypoint.sh file which contains the following:

#!/bin/bash
cd /app/stable-diffusion
nohup bash -c "lt --port 8080 &" && sleep 4
conda run -n ldm python webui.py --gfpgan-dir GFPGAN --no-progressbar-hiding 

Does it work on your side?