Why use a public IP?
Public IP addresses are useful for accessing your Gradient job container from the outside world. For example, you might want to run TensorBoard, a simple flask web server, or even SSH directly into your running job instance to see what is going on.
Gradient Notebooks and Jobs automatically have public IP addresses. For Jobs, you can additionally pass in which ports you would like to open on the running Job.
Opening Ports On Your Job
You can easily open ports in to your Gradient job using the --ports syntax.
If you are familiar with the Docker syntax for port forwarding, it is helpful to note that we use a similar form of:
For example, if you are running TensorBoard within your image and want to expose port 8888 to the world, you would just pass the following:
paperspace jobs create --container tensorflow/tensorflow:latest --command './run.sh' --ports 8888:8888
Within the interface you can see that this job has the public IP address of 18.104.22.168, so now any traffic that you send through 22.214.171.124:8888 will get redirected to the job's container on that same port.
Considerations / Additional Notes
Public IP addresses are not persistent. This means that once a job terminates, the public IP address is released back into the pool and is no longer available.