Manual Docker Configuration

If you are familiar with Docker, you can build your own Docker image instead of having Gradescope build your container image from a zip file. This can provide more control and speed up build times by taking advantage of local Docker build caching. It is not necessary to use this option to run your autograders on Gradescope, but it is available as an advanced feature.

Docker documentation

The rest of this page assumes familiarity with the Docker platform, including building a Docker image and pushing it to a container registry such as DockerHub. For more information, please see the official Docker documentation.


To use the "Manual Docker Configuration" option, you'll need to use gradescope/autograder-base as the base image for your Docker image. You can also other operating systems such as fedora, or different versions of Ubuntu - see the complete list on DockerHub.

Note: You can now also use different base images with a zip file upload.

You'll need to ensure that your image contains the run_autograder script at the path /autograder/run_autograder. This should match the requirements described on the specifications page. In particular, at the end of the script, the results should be in /autograder/results/results.json with the correct formatting.

Any setup can be done in the Dockerfile, so there is no need for a script. You can use one if it's easier though.

Beyond this, there are no other requirements on the structure of your Docker image, so you can organize it as you wish.

If you're just getting started, you can look at our sample Dockerfile. This example puts all the autograder source in a source directory and uses a file similar to the zip file upload method, so it can be a good transition path for going from a zip file to fully custom Docker builds.

You may also wish to refer to the Dockerfile reference docs.

Private Docker Hub Repositories

If your Docker Hub repository is private, you'll need to give the user gradescopeecs access to your repository. You may want to do this to make sure that students cannot download your autograder image.

Running autograder images locally

To run your autograder image locally, you will currently need to bypass our autograder harness because otherwise it will try to communicate with Gradescope by default. You can do this by mounting a sample submission into the /autograder/submission directory and then running /autograder/run_autograder directly. Here's an example command; replace the path to the submission, results directory, and Docker image name with the appropriate values.

docker run --rm -v /path/to/submission:/autograder/submission -v /path/to/results:/autograder/results username/image_name:tag /autograder/run_autograder && cat /path/to/results/results.json

or to start an interactive session:

docker run --rm -it -v /path/to/submission:/autograder/submission -v /path/to/results:/autograder/results username/image_name:tag bash

Minor notes:

  • --rm is added to clean up the container after it exits. You can remove it if you want to inspect container logs or state afterwards.
  • The /autograder/results directory should be mounted to a path on your host so that you can inspect the results.json file that your autograder produces.