🚀 Installation

We provide two installation paths. Docker is recommended — the image ships with UCX, flash-attn, sglang, and CUDA prebuilt.

🐳 Option A: Docker (recommended)

1. Pull the image

docker pull frankleeeee/sglang-omni:dev

2. Run the container

docker run -it \
    --shm-size 32g \
    --gpus all \
    --ipc host \
    --network host \
    --privileged \
    frankleeeee/sglang-omni:dev \
    /bin/zsh

3. Install sglang-omni inside the container

git clone git@github.com:sgl-project/sglang-omni.git
cd sglang-omni

uv venv .venv -p 3.12
source .venv/bin/activate

uv pip install -v -e .   # drop `-e` for a non-editable install

🛠️ Option B: Manual install

Build the prerequisites first:

• UCX 1.20.x with CUDA + verbs support — follow upstream, or reuse the exact build flags in docker/Dockerfile.

• flash-attn — install a prebuilt wheel matching your torch==2.9.1 / CUDA version.

Then install:

git clone git@github.com:sgl-project/sglang-omni.git
cd sglang-omni

uv venv .venv -p 3.12
source .venv/bin/activate

uv pip install -v -e .   # drop `-e` for a non-editable install