Frlai/improve leapp export

[frlai]

Description

Small additive updates for leapp.

1. Pins the leapp version to latest 0.6.1 which improves robustness for slicing operations. Previously this wasn't supported but now new features should automatically enable more types of operations.
2. Adds graph level expected frequency to leapp yaml. This change helps downstream deployment set default frequency based on training configs.
3. Adds the Isaac Lab env.yaml to the exported leapp bundle for convenience. Isaac sim 6.2 will add leapp controller features. the env.yaml is required to set things up. This will make porting the policy and its environment much more convenient.

Type of change

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

Release backport

• Backport this pull request to the active release branch after it merges into develop

Checklist

• I have read and understood the contribution guidelines
• 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 added a changelog fragment under source/<pkg>/changelog.d/ for every touched package (do not edit CHANGELOG.rst or bump extension.toml — CI handles that)
• I have added my name to the CONTRIBUTORS.md or my name already exists there

[greptile-apps]

Greptile Summary

This PR upgrades the optional LEAPP dependency to 0.6.1 and enriches exports with policy-frequency metadata and the originating environment configuration.

• Computes graph frequency from the environment’s simulation timestep and decimation.
• Applies the metadata and env.yaml export consistently across RL-Games, RSL-RL, SB3, and SKRL.
• Extends direct and multi-backend export-flow tests to validate the new artifacts.
• Updates deployment documentation, changelog, and dependency lock data.

Confidence Score: 5/5

The PR appears safe to merge, with no concrete changed-code failure identified.

The frequency calculation matches the canonical policy step period for both supported environment families, all exporters apply the same artifact contract, and the updated integration tests validate metadata-to-environment parity.

Important Files Changed

Filename	Overview
scripts/reinforcement_learning/leapp/export_utils.py	Adds a shared graph-configuration helper using the canonical environment policy period.
scripts/reinforcement_learning/leapp/rl_games/export.py	Adds graph frequency metadata and environment YAML to RL-Games exports.
scripts/reinforcement_learning/leapp/rsl_rl/export.py	Adds the shared metadata and environment artifact to manager-based and direct RSL-RL exports.
scripts/reinforcement_learning/leapp/sb3/export.py	Adds graph frequency metadata and environment YAML to SB3 exports.
scripts/reinforcement_learning/leapp/skrl/export.py	Adds graph frequency metadata and environment YAML to SKRL exports.
source/isaaclab_rl/test/export/test_leapp_export_flow.py	Verifies graph frequency and environment configuration parity across backend export flows.
source/isaaclab_rl/test/export/test_rsl_rl_direct_export_flow.py	Extends direct-environment export coverage for the new metadata and environment artifact.
pyproject.toml	Pins the optional LEAPP dependency to version 0.6.1.
uv.lock	Resolves LEAPP 0.6.1 and refreshes associated lock metadata and platform artifacts.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  C[Environment configuration] --> F[Compute frequency from sim.dt × decimation]
  F --> G[LEAPP GraphConfigs]
  P[Policy checkpoint] --> E[RL-specific exporter]
  G --> E
  E --> L[Compile LEAPP graph]
  L --> A[Model and graph artifacts]
  C --> Y[Write env.yaml]
  Y --> A
  A --> D[Downstream deployment]

Loading

_{Reviews (1): Last reviewed commit: "bumped pinned leapp version to 0.6.1, le..." | Re-trigger Greptile}

[isaaclab-review-bot]

Isaac Lab Review Bot

Reviewed the LEAPP 0.6.1 integration, including policy-frequency metadata, bundled environment configuration, all four RL export backends, documentation, release metadata, and export-flow tests. The changed producer and artifact paths are consistent across backends.

• Design and architecture: Policy frequency is derived centrally from the environment control period (sim.dt * decimation) and passed through the existing deferred-import structure used by the RSL-RL, RL-Games, skrl, and SB3 exporters. Each backend writes env.yaml into the same graph-specific artifact directory, keeping the bundle layout uniform.
• API: No existing Isaac Lab CLI option or export artifact is removed or renamed. The LEAPP dependency is pinned to 0.6.1 to match the newly used graph configuration API, while the new frequency metadata and env.yaml artifact are documented and included in the package changelog.
• Implementation: The graph configuration and environment dump are wired into all four compile paths using the same env_cfg, save_path, and graph_name. Integration assertions verify that both artifacts exist and that the graph frequency matches 1 / (sim.dt * decimation), including the direct-environment RSL-RL path.

No blocking issues. No inline issue met the actionable-evidence threshold; the assessment above records the review feedback.

Automated review; human maintainers own approval decisions.

[kellyguo11]

@StafaH do you mind checking on the uv.lock changes?

[StafaH]

Thanks for the PR @frlai. I left a comment about moving the logic in a follow up PR, but on that note, we have also been migrating everyone to move logic to uv run isaaclab. Currently you can do:

uv run isaaclab train <rest of args>
uv run isaaclab benchmark startup <rest of args>
uv run isaaclab teleop record <rest of args>

Might be time to add leapp export, wdyt?

uv run isaaclab export <rest of args>

This would require 1) moving logic to isaaclab_rl and 2) getting hooked up correctly to cli.py

[StafaH]

Why is the lock changing? Optional dependencies usually don't change the lock that much, you can double check if this lock change is necessary locally by doing a uv lock --check on a fresh install (no leapp installed just base install)

[frlai]

Done. Previous issue was that I didn't clean the lock file before pushing. newest version has exactly 3 changes.

[StafaH]

The changes look reasonable, but just FYI, we moved almost all logic out of the scripts into isaaclab_rl, so leapp is currently the only logic left behind here. Benefit of moving to isaaclab_rl is that downstream package install gets access.

[frlai]

yeah I would prefer that too. previously noticed that we were doing that but wasn't sure about the scope. I think it's relevant enough to add to this PR. Unless you have a preference for smaller PRs.

[frlai]

@StafaH if this looks good can you approve it? we can discuss the migration it isaaclab_rl separately.

[frlai]

run-ci