Adds record at close functionality to the RecorderManager

[jtigue-bdai]

Description

Introducing a close function to the recorder manager which exports the data to file when the environment is closed and closes the recorder terms.

Fixes # (issue)

Type of change

• New feature (non-breaking change which adds functionality)

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

[ooctipus]

LGTM!

[kellyguo11]

@peterd-NV do you mind taking a quick look at this change to make sure it's ok for mimic?

[peterd-NV]

Hi @kellyguo11 @jtigue-bdai, I just have a couple of questions to make sure that everything will work with Mimic workflows:

1. The new recorder manager close function will always be called during env.close(). This will result in whatever existing episode data for each env to be written to the dataset file. This will not overwrite any previously written data/episodes to the dataset file, correct?
2. The current episode when recording manager close is called may not be a complete episode. For example, when doing multi-env data generation, there may be envs in the middle of an episode when the script is shutdown. In these cases, we do not want these episodes written to file. Can we continue to prevent the close function from writing the remaining incomplete episodes to file by using the episode_succeeded mechanism in the export_epsiodes function of the recorder manager?

[jtigue-bdai]

Hi @kellyguo11 @jtigue-bdai, I just have a couple of questions to make sure that everything will work with Mimic workflows:

1. The new recorder manager close function will always be called during env.close(). This will result in whatever existing episode data for each env to be written to the dataset file. This will not overwrite any previously written data/episodes to the dataset file, correct?
2. The current episode when recording manager close is called may not be a complete episode. For example, when doing multi-env data generation, there may be envs in the middle of an episode when the script is shutdown. In these cases, we do not want these episodes written to file. Can we continue to prevent the close function from writing the remaining incomplete episodes to file by using the episode_succeeded mechanism in the export_epsiodes function of the recorder manager?

Thanks @peterd-NV

1. Correct this shouldn't affect any previous episode data.
2. I believe it should still follow the same rules for recording as the original so the episode_succeeded should still function but what I don't know is if the success_results will be set if not entering the record_pre_reset. We could add a cfg parameter export_in_close to be able to toggle this though. Similar to the export_in_record_pre_reset.

[peterd-NV]

Thanks for the answers @jtigue-bdai.

Yes, if we could add a cfg parameter to toggle if export occurs in close that would be great. Ideally, if we could have this parameter default to be False that may be the safest option for existing Mimic workflow users. I believe most Mimic users are running parallel data generation/recording so having export in close default to False will guarantee their generated datasets don't contain unwanted incomplete demos.