Unsloth Docker on WSL2 (Debian Trixie)

1. Add the NVIDIA Container Toolkit repo

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
  sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

2. Install the toolkit

export NVIDIA_CONTAINER_TOOLKIT_VERSION=1.17.8-1
sudo apt-get update && sudo apt-get install -y \
  nvidia-container-toolkit=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  nvidia-container-toolkit-base=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  libnvidia-container-tools=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  libnvidia-container1=${NVIDIA_CONTAINER_TOOLKIT_VERSION}

If apt throws a signature error, add --allow-insecure-repositories to the apt-get update call. This is a known Debian Trixie issue with NVIDIA’s SHA1-signed repo as of Feb 2026.

3. Configure Docker to use the NVIDIA runtime

sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

4. Verify GPU passthrough

docker run --rm --gpus all nvidia/cuda:12.3.0-base-ubuntu22.04 nvidia-smi

You should see your GPU listed with full VRAM. If this fails, stop here - the container won’t see the GPU either.

5. Stop any running containers

docker stop $(docker ps -q)

6. Run Unsloth

docker run -d \
  -e JUPYTER_PASSWORD="yourpassword" \
  -e JUPYTER_PORT=8888 \
  -p 8888:8888 -p 2223:22 \
  -v $(pwd)/work:/workspace/work \
  --gpus all \
  unsloth/unsloth

Port 2222 may be in use by another service. 2223 is the safe alternative for SSH.
Change the password.

7. Access Jupyter Lab

Open in your Windows browser: http://localhost:8888

Fix (2026-03-13)

pip install --upgrade --force-reinstall --no-cache-dir --no-deps unsloth unsloth_zoo

pip install "numpy<2.0" --force-reinstall

import unsloth
print(unsloth.__version__)

These are for allowing the traefik routing / LAN access

netsh interface portproxy add v4tov4 listenport=8888 listenaddress=0.0.0.0 connectport=8888 connectaddress=172.31.81.80


netsh advfirewall firewall add rule name="WSL2 Jupyter 8888" dir=in action=allow protocol=TCP localport=8888

Notes

• There is only one Unsloth image: unsloth/unsloth. Works for all current GPU generations including Blackwell/50-series.
• Always mount a work volume or your files disappear when the container stops.
• The container runs as a non-root unsloth user. Use USER_PASSWORD env var if you need sudo inside the container.
• SSH into the container (optional): ssh -i ~/.ssh/container_key -p 2223 unsloth@localhost