A minimal CUDA and cuDNN install on top of the official python:3.x-slim base image.
- β
Starts from the official
python:3.x-slimbase image - π Adds a single
micromambaexecutable to install CUDA and cuDNN - 𧬠Matrix build for Python {3.8, 3.9, 3.10, 3.11, 3.12, 3.13}, CUDA {11.8}, and cuDNN {8.9}
- π¦ Multi-platform build for
linux/amd64andlinux/arm64 - π Image size is only 1.8GB
A matrix of tags are available that follow the format superlinear/python-gpu:$PYTHON_VERSION-cuda$CUDA_VERSION, see the Docker Hub repository for a full list.
docker run -it --rm superlinear/python-gpu:3.11-cuda11.8 /bin/bash
FROM superlinear/python-gpu:3.11-cuda11.8
...