Problem running the script in container

Hi, I’ve some issues/questions.

  1. Does the container has “access” to the internet? There are huggingface pretrained LM used in the script. So do we need to download it before, save it to a cache dir, include this cache dir in the image before submitting it, or can we just call the script and let it download the LM from the container?

  2. I copied the Dockerfile from and make minimal changes (like deleting the RUN ap-get update part and copying my scripts instead of the baseline scripts). The baseline image would work fine, but the container from my image keep getting errors.
    First I get permission denied error for the running script, then I changed the Dockerfile by adding

RUN chmod +x /

but still then get the error: / No such file or directory

How to fix this?

Thanks for the help.

Dear Dirk,

Thank you for reaching out and sorry for not being clear enough.
The container has no access to the internet, to ensure that the software is fully installed in the container (and that the test data can not be leaked while the task is running, but I clearly second that). This way, we hope that we can increase the replicability/reproducibility of software submissions because everything is already installed in the image (in case some services are not available or incompatible in a few years).

So please add your models to the image, that will resolve the problem.

For your second problem: the error that you describe / No such file or directory.
I remember that such error messages can happen when the file is missing or when the shebang is missing or wrong. Because the file is already there (as you can run chmod +x /, I think it must be the shebang.

Does your file / start with something like this in the first line?


Adding such a shebang as first line should resolve it.
(In case you have further problems or the steps above do not resolve it, please do not hesitate to let me know)

Thanks for participating!

Best Regards,