Description

This PR adds Ray support, which enables a lot of really cool stuff by leveraging the existing Hydra support, including but not limited to:

• Several training runs at once in parallel or consecutively with minimal interaction
• Using the same training setup everywhere (on cloud and local) with minimal overhead
• Tuning hyperparameters
• Tuning hyperparameters in parallel on multiple GPUs and/or multiple GPU Nodes
• Simultaneously tuning model hyperparameters for different environments/agents
• Resource Isolation

I know this PR seems huge, but most of the code diff is config files / argparser stuff / documentation / comments

My main project at BDAI is changing from RL to LfD effective November 1st, so I'm posting this PR as early as possible so I have as much time as possible to address comments.

It would be much appreciated if the NVIDIA folks can work with me to get this reviewed ASAP. I realize that this is a pretty big PR; but I also think that it adds a lot of cool functionality, and the merging process will go much smoother if I am able to devote time to this while at work. Thanks! ;)

Fixes # (issue#1190), (issue#1213)

Type of change

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

Screenshots

Checklist

• I have run the pre-commit checks with ./isaaclab.sh --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have updated the changelog and the corresponding version in the extension's config/extension.toml file
• I have added my name to the CONTRIBUTORS.md or my name already exists there