Docker Setup for Applio with CUDA Support Not Working on unRAID

[DEV-MalletteS]

Description: I’m trying to set up Applio using Docker with CUDA support, but I’m unable to make it work. The repository includes a Dockerfile and docker-compose.yml, and it’s typically hosted on Gradio. I believe the setup shouldn’t be too complicated, but I’m facing issues.

Steps I’ve Taken:

1. Environment Setup:

   • Added GRADIO_SERVER_NAME=0.0.0.0 to the .env file, similar to other Gradio-based hosted servers (e.g., FaceFusion).
   • Exposed port 6969 in the Docker configuration.

2. Dockerfile Configuration:

   • Set WORKDIR to /app.
   • Used python:3.10-bullseye as the base image.
   • Created a Python virtual environment in /app/.venv.
   • Added /app/logs/ as a volume.
   • Updated the PATH to include the virtual environment.

3. CUDA Support:

   • Added the following parameters for CUDA support:

     --runtime=nvidia
     --gpus=all
     

   • Set the following environment variables:

     NVIDIA_DRIVER_CAPABILITIES=all
     NVIDIA_VISIBLE_DEVICES=all
     

Expected Behavior:

The Docker container should start successfully, and the Gradio server should be accessible on port 6969 with CUDA support enabled.

Actual Behavior:

The setup does not work as expected, and I’m unable to access the Gradio server.

Additional Context:

• Repository: IAHispano/Applio
• The project relies on CUDA, so proper GPU support is essential.
• I’ve followed similar setups for other Gradio-based projects (e.g., FaceFusion), but this one isn’t working.

Request:

Could someone help me troubleshoot this issue? Specifically:

1. Are there any missing steps in the Docker or CUDA configuration?
2. Are there any known issues with the Dockerfile or docker-compose.yml in the repository?
3. Are there additional logs or debugging steps I should provide?

Thank you!

---

Code Snippets:

Dockerfile:

WORKDIR /app

FROM python:3.10-bullseye

RUN python3 -m venv /app/.venv && \
    (Python environment is included as a runtime folder)

VOLUME ["/app/logs/"]

ENV PATH="/app/.venv/bin:$PATH"

Environment Variables:

GRADIO_SERVER_NAME=0.0.0.0
NVIDIA_DRIVER_CAPABILITIES=all
NVIDIA_VISIBLE_DEVICES=all

[blaisewf]

Did you find any way to fix it? pr would be appreciated, we don't actually use docker