This repo aims at prepare the online coding experience for both Kubeflow and docker compose.
Just use the sharockys/pytorch-notebook
and sharockys/vscode-py-cuda
in Kubeflow Jupyter Server to spawn a container, and everything will be ready to use.
docker run --gpus all -it --rm -p 8888:8888 -v $(pwd):/workspace sharockys/cuda-notebook
git clone https://github.com/haoxian-lab/kubeflow-web-ide
cd kubeflow-web-ide/docker-compose-files/notebook
UID="$(id -u)" GID="$(id -g)" docker-compose up -d
git clone https://github.com/haoxian-lab/kubeflow-web-ide
cd kubeflow-web-ide/docker-compose-files/code-server
docker-compose up -d
Then add code-server.localhost $IP_ADDRESS
to your /etc/hosts
. Alternatively, you can set up a DNS record for it.
Then with https://code-server.localhost
in web browser, you will have the full functioning code server in your remote server.
It's better to have TLS for using Jupyter Notebook in Code Server. So in the docker-compose.yml
for code-server, a Caddy Reverse Proxy is set up with self-signed cert.
This is convenient for boosting coding experience on a shared server where we use only have access to Docker but no sudo permission.
https://github.com/Tverous/pytorch-notebook https://github.com/huggingface/transformers/blob/main/docker/transformers-all-latest-gpu/Dockerfile