Demo: Stable-Audio-Open-1.0 using Gradio

https://huggingface.co/stabilityai/stable-audio-open-1.0

https://www.gradio.app/

The Gradio code is configured to listen on an IPv6 port for running on SaladCloud; use IPv4 for local runs and tests:

demo.launch( server_name="[::]", server_port=8000 ), for IPv6 usage

demo.launch( server_name="0.0.0.0", server_port=8000 ), for IPv4 usage

Deploy on SaladCloud

Create a container group with the following paramaters and access the app using the generated Access Domian Name:

Image Source: saladtechnologies/richardx:0.0.3-stable (our prebuilt image or yours)

Replica Count: 1 (can only be one for this UI app)

Resource：2 vCPU, 12 GB Memory, any GPU types with 16 GB or more VRAM

Container Gateway: Enbaled, Port 8000

Environment Variables: HF_TOKEN (YOUR_HUGGINGFACE_ACCESS_TOKEN_READ)

The model weights (around 6 GB) will be dynamically downloaded from Hugging Face when the container is running.

Local Run

Build & Push

docker image build -t YOUR_IMAGE_NAME -f Dockerfile .

docker login

docker push YOUR_IMAGE_NAME

Run

docker run -it --rm -p 8000:8000 --gpus all
-e HF_TOKEN="YOUR_HUGGINGFACE_ACCESS_TOKEN_READ"
YOUR_IMAGE_NAME

Run with the pre-downloaded model weights on the host

docker run -it --rm -p 8000:8000 --gpus all
-v /home/CURRENT_USER/.cache/huggingface:/root/.cache/huggingface
-e HF_TOKEN="HUGGINGFACE_ACCESS_TOKEN_READ"
YOUR_IMAGE_NAME