FAQ

How do I run my run bundle on a GPU?

See our article on running jobs that use GPUs.

Why is my run bundle in staged for a long time?

The staged state means that the bundle is ready to run but it is waiting for appropriate resources to free up. CodaLab has a limited set of machines available (we're offering it for free after all), so when there are a lot of people wanting to run jobs, you might have to wait a long time. You can check out the CodaLab status page to see where your jobs are in the queue. If you'd like to run your job earlier, you can easily attach your own compute workers, which you might want to do anyway if you have fancier GPUs, say.

How do I share a bundle only with other CodaLab users?

By default, all bundles (and worksheets) on CodaLab are public (and we encourage this). But if you need to make your bundle private to you and your team, check out the instructions here.

I received an ImportError because I don't have the appropriate library installed. How do I install libraries in CodaLab?

CodaLab runs all run bundles in Docker containers. See our article about specifying Docker containers. You can see how to make your own Docker images.

How do I check how much space my bundles are using?

You can use cl uinfo to find out how much disk space you have left. To find out which bundles are taking the most space, you can check "My Dashboard" or type cl search .mine size=.sort-.

How do I reduce the amount of disk usage?

If you have used up your disk quota and want to reduce your disk usage, here are a few tips:

• Make sure that you don't have unused bundles lying around. Type cl search .mine size=.sort- to find the largest bundles. Note that some of these bundles might be floating (not showing up on any worksheet).
• If you have intermediate bundles whose contents you don't need, you can use cl rm -d <bundle> to delete only the contents of the bundle but keep the metadata. This way, you keep the provenance and can regenerate this bundle in the future.
• If there are selected files that you wish to keep in a bundle, you can use cl make <bundle>/<important-file> to make a new bundle that only has a copy of the desired file and then you can do cl rm -d <bundle> to remove all the other files in that bundle.

Common Errors

I upgraded CodaLab and now I'm getting an error like: ValueError: unknown url type: local/rest/worksheets?specs=%2F

Run cl alias main https://worksheets.codalab.org and retry your command.

When I try to execute a command like make on CodaLab, it fails because of "read-only" permissions. What do I do?

Each bundle uploaded to CodaLab is read-only. If a command you are running creates new files, you'll have to create a duplicate of the in the run command. For example, instead of running cl run :src 'cd src && make, run cl run src-orig:src 'cp -RL src-orig src && cd src && make'. Your freshly compiled executable will now be present in the src directory inside the last bundle.