Gradient Job Runner - Get logs in local console


#1

Hi,

I’m using the Gradient Job Runner with the paperspace-python cli to train my NN. When I start a Job it uploads all the stuff and creates a new Job.The only thing that bothers me is that I always have to open the paperspace website, click on the newly created Job and observe the logs in the browser to see if everything is working as expected. My local console just logs … “Finished creating archive: blabla … Job created - ID: sdoi32q8rdsjr”.
I would love if the connection to the server would stay active, log everything my script logs via ssh in my local console and disconnects when the jobs is finished. Is there a convenient to get this working?

EDIT: I found that I can get the job logs by typing paperspace-python jobs logs --jobId sdoi32q8rdsjr, so I change the question: Is there a way to give the create command a parameter so it gets the logs automatically without needing a second command with the specific jobid?

EDIT2: Solved it with paperspace-node

Thanks in advance!


#2

Hi @malte2408 This is a great question. We recently introduced a whole new CLI which automatically streams the logs to your terminal. You can install the CLI with:

pip install -U gradient

(We recommend installing in a virtual environment)

More info here: https://docs.paperspace.com/gradient/get-started/install-the-cli

We still offer Jobs but we’re migrating to a new concept called Experiments which behave like Jobs but also enable you to run distributed training and hyperparameter search where an Experiment will contain multiple jobs.

The syntax for creating a Job from an Experiment is:

gradient experiments run singlenode ...

Here is some more info: https://docs.paperspace.com/gradient/experiments/run-experiments-cli

Hope that helps.