generative join

Setup Instructions

Approximate setup time: 20 minutes. (5 minutes of work + 15 minutes of installation in the background.)

Step 1: Install the gcloud CLI

First, you'll need to install the Google Cloud command line tools.

• Follow the instructions on the installation page
• run gcloud init as described in this guide and configure the tool with the ID of your new Cloud project.

Step 2: Connect to the configured gcloud instance

We've created a Google Cloud instance with an attached GPU for you to use with this development workflow. Run the following snippet to set a few environment variables:

export ZONE="us-central1-a"
export PROJECT="probcomp-caliban"
export INSTANCE_NAME="generative-join"

Configure the ssh credentials required to reach the instance with the gcloud cli:

gcloud compute config-ssh --project $PROJECT

You should now be able to log in to the instance with the following command:

ssh $INSTANCE_NAME.$ZONE.$PROJECT

If you have any trouble, visit the cloud console and make sure that the generative-join instance is running. If it's not, select the instance's checkbox and click "Start" to boot the machine up again for use. After a couple minutes, the instance should be running and the ssh command should work.

Step 3: Set up git access and clone the repository

Once you are connected to the VM, add a github ssh key to the VM, so that you can clone the repository (and push changes back to it) from the machine.

1. run `ssh-keygen -t ed25519 -C "[email protected]" (filling in your email address). This will generate an SSH key. I recommend you use a passphrase if you are using a cloud machine that probcomp gave you access to (since the probcomp members who set up the machine for you have access to it too).
2. Run eval $(ssh-agent -s) and then ssh-add ~/.ssh/id_ed25519 to add the key to the ssh agent.
3. Run cat ~/.ssh/id_ed25519.pub and copy the output to your clipboard.
4. Visit the github ssh key settings and add the key to your account. (Click "New SSH Key" in the top right corner, and copy in the public key you just generated.)

Now you can clone the repository:

git clone [email protected]:probcomp/generative_join.git

Step 4: Connect to VS Code

To run the code, we have set up a development VM which is pre-configured to load in the dependencies needed to run Bayes3D. This VM will run on the google cloud server (and we can later configure it to run on local hardware). We have set this up so that you can connect to this VM directly from VS Code on your desktop, to develop and run code through it.

Start VS code on your desktop. If you don't have it already, install the Dev Containers VS Code extension, by Microsoft. (You can do this by clicking the Extensions icon on the left, which looks like three connected squares with a disconnected square in the top right. In the Extensions tab, search for "Dev Containers".)

Once you have Dev Containers installed, click on the >< icon on the bottom left corner of VS Code to get the Connect to Host menu. Choose your VM, which should be among the choices due to using config-ssh, and connect. Once on your cloud VM, open the generative_join folder you created by cloning the git repository.

Step 5: Run the code through VS Code

At this point, you should be able to run the code in the generative_join through your VS Code window.

To run the tutorial notebook, open it and run the cells!

You are welcome push changes to the code to branches on github.

Step 6: Shut down the VM

When you're done working, visit the cloud console, select the generative-join image and hit the "Stop" button that appears in the pop-up menu.

When you're ready to work again, visit the cloud console and hit "Start" to boot the machine up again for use.