KubeflowPipelineMNIST

Local Jupyter Lap Setup

0. OPTIONAL : If you don't have docker installed, you can install it by running ./install-docker.sh

GPU

1. ./setup-nvidia-container-toolkit.sh
2. ./build-local-notebook-server.sh
3. ./start-local-notebook-server-gpu.sh
4. 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).
5. Inside jupyter lab, click File->New->Terminal
6. Clone this git repo (eg: git clone https://github.com/AlexandreBrown/KubeflowPipelineMNIST.git)

CPU

1. ./build-local-notebook-server.sh
2. ./start-local-notebook-server-cpu.sh
3. 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).
4. Inside jupyter lab, click File->New->Terminal
5. 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