- We are going to use MiniKF: https://aws.amazon.com/marketplace/pp/prodview-7shm7yqkubjhg
- You can ignore the red permission issues at the top
- Launch an instance using the EC2 console
- Make sure to enable public ip
- Open ports 22, 80, 443, 8443, 32123
- ssh into the machine and run minikf to see the progress
- Wait for it to finish and log into kubeflow in the address and the credentials provided
- Make sure to allow self signed certificates
- Create a new volume called
models - Create a notebook instance that uses that volume
- Open a terminal in the notebook instance and clone the repo
- Create a data directory and download data
wget https://www.robots.ox.ac.uk/~vgg/data/pets/data/images.tar.gz
wget https://www.robots.ox.ac.uk/~vgg/data/pets/data/annotations.tar.gz
tar -xzf images.tar.gz
tar -xzf annotations.tar.gz
- Install requirements
- Open the training notebook Training.ipynb
- Train and save the model on the
modelsvolume - ssh back into the machine and deploy the model
kubectl apply -f K8s/
- Make sure the deployment is running
kubectl get pods -n kubeflow-user
- Check the model with the Inference.ipynb notebook
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent