Stable Diffusion WebUI Forge - Neo

This repository contains a personal, modified fork of ai-dock/stable-diffusion-webui-forge, adapted to run Haoming02's Forge Neo.

Important

This is an independently maintained derivative hosted at gabszap/stable-diffusion-webui-forge. It is not an official AI-Dock image.

The project retains AI-Dock's branding, container structure, and runtime integrations while replacing the original Forge installation with Forge Neo.

Quick Facts

Item	Detail
Image Tag	ghcr.io/gabszap/stable-diffusion-webui-forge:neo-2.28.1-py3.12-v3-cuda-13.0.3-24.04
Forge Release	2.28.1
Source Commit	609d3e3
Python	3.12
Torch / CUDA	Torch 2.13+cu130 / CUDA 13.0.3
Compatibility	NVIDIA GPUs only

Requirements

• Linux amd64 host.
• NVIDIA GPU with driver supporting CUDA 13.0 (R580+ driver family recommended).
• Sufficient VRAM for your checkpoints. (12GB+ Recommended)

Vast.ai Setup

A preconfigured Vast.ai template is available below:

Open the Vast.ai template

Provisioning

Note

Models and third-party configurations are not bundled with this image.

Use a provisioning script to configure the container automatically. Example scripts are available in config/provisioning.

Provisioning scripts can be used to automatically install or download:

• Models and VAEs
• Extensions
• LoRAs and embeddings
• Custom configuration files
• Additional system or Python packages

Local Docker Run

Create a persistent workspace directory:

mkdir -p workspace

Start the container:

docker run -d \
  --name forge \
  --restart unless-stopped \
  --gpus all \
  -p 7860:7860 \
  -e WEB_ENABLE_AUTH=true \
  -e WEB_USER=user \
  -e WEB_PASSWORD=your-secure-password \
  -e FORGE_ARGS=--api \
  -e WORKSPACE=/workspace \
  -v "$PWD/workspace:/workspace:rshared" \
  ghcr.io/gabszap/stable-diffusion-webui-forge:neo-2.28.1-py3.12-v3-cuda-13.0.3-24.04

Open the WebUI at:

http://localhost:7860

View the container logs:

docker logs -f forge

Stop the container:

docker stop forge

Start it again:

docker start forge

Stop and remove it:

docker rm -f forge

Common Controls

Check the Forge service status from inside the container:

supervisorctl status forge

Restart the Forge service:

supervisorctl restart forge

Configure Forge startup arguments with the FORGE_ARGS environment variable:

-e FORGE_ARGS="--api --listen"

Remote Access Notes

OpenCode TUI over SSH

Running OpenCode's interactive TUI directly through an SSH session can feel noticeably laggy when network latency is high, particularly when the host is geographically distant.

This is caused by terminal round-trip latency and does not indicate a Forge, GPU, or container performance issue.

For the most responsive experience:

• Run the OpenCode TUI locally and connect it to the backend running on the host using opencode attach.
• Use OpenCode's web interface.
• When neither option is suitable, choose a geographically closer host or a lower-latency connection.

Known Issues

Expected Gradio and Pillow dependency warning

Gradio 4.40.0 declares Pillow >=8,<11, while Forge Neo 2.28.1 pins Pillow 12.3.0. Package validation may therefore report a dependency conflict.

The image follows Forge Neo's Pillow requirement and has been tested with this combination. Do not manually downgrade Pillow solely to suppress the warning.

Attribution, Licensing, and Funding

This image is a derivative work. Please see the following files for original licensing and notices:

• LICENSE.md
• NOTICE.md

Support the original AI-Dock maintainers:

• Support AI-Dock
• Support Rob Ballantyne

AI-Dock-originated code remains subject to AI-Dock’s Custom Software License. Forge Neo and other third-party components remain subject to their respective licenses. No additional permissions are claimed.