OPTIONAL : If you don't have docker installed, you can install it by running ./install-docker.sh
GPU
./setup-nvidia-container-toolkit.sh
./build-local-notebook-server.sh
./start-local-notebook-server-gpu.sh
Access jupyter lab via http://127.0.0.1:8889/lab?token=YOUR_TOKEN where YOUR_TOKEN is the token created in step 3. (check the output from step 3 to know your token).
Inside jupyter lab, click File->New->Terminal
Clone this git repo (eg: git clone https://github.com/AlexandreBrown/KubeflowPipelineMNIST.git)
CPU
./build-local-notebook-server.sh
./start-local-notebook-server-cpu.sh
Access jupyter lab via http://127.0.0.1:8889/lab?token=YOUR_TOKEN where YOUR_TOKEN is the token created in step 2. (check the output from step 2 to know your token).
Inside jupyter lab, click File->New->Terminal
Clone this git repo (eg: git clone https://github.com/AlexandreBrown/KubeflowPipelineMNIST.git)
Local Setup Result
Pipeline
Execute the cells from the notebook (see /notebooks)
Upload the generated pipeline yaml file to the Kubeflow pipeline UI
Pipeline Result
About
Kubeflow Pipeline sample that uses custom docker image with private python code inside the image as base image for the pipeline components.