Noisy warning while using the DeepHyper CLI

[Deathn0t]

I quote @felker from #23

Still, curious that DeepHyper loads TensorFlow (and its noisy warnings) even for the --help dialog.

The problem is happening when using:

$ deephyper --help

Indeed these warnings are flooding the shell and are very annoying, I create this issue so that we keep thinking about solving this.

[felker]

For my project, I just suppress them in __init__.py wiht:

warnings.filterwarnings('ignore',
                        category=FutureWarning,
                        message=r"passing \(type, 1\) or '1type' as a synonym of type is deprecated",
                        module="tensor*")

They come from TensorFlow with NumPy >= v1.17.0.

But it is probably better to bypass importing TensorFlow entirely when --help is used, if possible.

[Deathn0t]

To build the DeepHyper CLI parser we import the search classes and do some introspection on them. So the PPO search using reinforcement learning is imported and it itself imports Tensorflow. I recently noticed that a Mpi4py mock was added to this process to avoid mpi errors while using the CLI when openmpi is not installed (e.g., Theta login nodes). Thus, the deephyper nas ppo.. was not working at all. Given this two issues, I decided to remove the PPO search from the CLI. Tensorflow is not loaded anymore while using the parser.