also worth noting, in the configuration tab of an experiment, before this issue all other args would be shown with their default value once the script would be executed. Now, only the args explicitely passed in the command line clearml-task are shown in this tab

this is caused by the remote instance installing clearml 1.1.1 which is incompatible with python3.11 due to the regex in the traceback. However, even when I specify to use a compatible clearml version in the requirements it installs clearml 1.1.1 anyways....

Im testing with the following script. It runs fine if its not part of a git repo. Placing it inside a git repo (though I am testing it in a private bitbucket in my work) and the args arent passed:

#script that just multiplies args, with one arg, x being obligatory
if __name__ == "__main__":
    import argparse

    parser = argparse.ArgumentParser()

    parser.add_argument("--x", type=float, required=True)
    parser.add_argument("--y", type=float, default = 2.)
    parser.add_argument...

"poetry_install_extra_args". I work with a git repo that has optional package dependency groups needed for training, so I need to install the packages with "poetry install --extras training". Using poetry_install_extra_args solves the issue. However, since this is a package dependant (some dont have the training dependencies, others have other dependency group names) fix, I would rather this not be fixed in the configuration vault, but something I can set via the pipeline code.