We currently support real HDFS by using pyarrow, but the problem with it is that it still requires libhdfs, which has to be installed separately and in some cases (on osx) is not at all trivial to install (brew's hadoop package doesn't provide it, for example). See iterative/dvc#4656 Also [pyarrow itself has caused us some problems](https://github.com/iterative/dvc/issues

Describe the bug
I found that some names agruments in framework aren't consistent.
So for example:

class SupervisedRunner(Runner):
    """Runner for experiments with supervised model."""

    _experiment_fn: Callable = SupervisedExperiment

    def __init__(
        self,
        model: Model = None,
        device: Device = None,
        input_key: Any = "features", 
      

Description
The submissions of the challenge hosts don't appear until they are made as a baseline but we are missing on the documentation for it.

Deliverables
Step by step guide on how to make a submission a baseline submission.

https://github.com/kislyuk/argcomplete allows for easy integration with bash auto-completion.

similar to #785 but for docker pull