To store the docker BUILD scripts of ONNX related docker images.
- onnx-base: Use published ONNX package from PyPi with minimal dependencies.
- onnx-dev: Build ONNX from source with minimal dependencies.
- onnx-ecosystem: Jupyter notebook environment for getting started quickly with ONNX models, ONNX converters, and inference using ONNX Runtime.
- onnxruntime-gpu-1.6-python: CUDA 10.2 / onnxruntime-gpu 1.6 docker for python.
- onnxruntime-gpu-1.7-python: CUDA 11.0.3 / onnxruntime-gpu 1.7 docker for python.
- onnxruntime-gpu-1.6-c: CUDA 10.2 / onnxruntime-gpu 1.6 docker for c.
- onnxruntime-gpu-1.7-c: CUDA 11.0.3 / onnxruntime-gpu 1.7 docker for c.
- Obtain the Docker images
You can clone this repository and build your desired image.
# onnx-base container
cd onnx-base
docker build . -t onnx-base
# onnx-developer container
cd onnx-dev
docker build . -t onnx-dev
# onnx-ecosystem container
cd onnx-ecosystem
docker build . -t onnx-ecosystem
# onnxruntime-gpu1.6-python container
cd onnxruntime-gpu1.6-python
docker build . -t onnxruntime-gpu1.6-python
Alternatively, you can pull a pre-built image from DockerHub.
docker pull onnx/onnx-base
docker pull onnx/onnx-dev
docker pull onnx/onnx-ecosystem
- Run the images
docker images
docker run -it onnx-base
docker run -it onnx-dev
docker run -p 8888:8888 onnx-ecosystem
docker run --gpus all -it --ipc=host onnxruntime-gpu1.6-python