Add v0.3.0 release notes - #1228
Conversation
 The documentation-only PR appears safe to merge with no outstanding actionable findings.
|
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
![arena-review-bot[bot]](https://avatars.githubusercontent.com/u/195794629?s=60&v=4){kind=small}
🤖 Isaac Lab-Arena Review BotRe-reviewed at SummaryDocs-only change adding a Findings🟡 🟡 Still open from earlier passes (threads are live, no need for new ones): Test CoverageNot applicable — documentation only. VerdictMinor fixes needed. Nothing here blocks the release; the CPU/Fabric limitation is the one I'd most want captured, since it changes the performance users get from a workflow these notes headline. |
|
|
||
| Isaac Lab-Arena is being developed as an open-source, shared evaluation framework that the community can | ||
| collectively enhance and expand. We invite you to try Isaac Lab-Arena 0.2 Alpha, share feedback, and help | ||
| collectively enhance and expand. We invite you to try Isaac Lab-Arena 0.3 Alpha, share feedback, and help |
There was a problem hiding this comment.
@sangeetas-nv @alexmillane do we still call 0.3 an Alpha release?
| Future releases will focus on VLM-augmented analysis to surface insights from large-scale evaluations, | ||
| Newton support for contact-rich rigid-body and deformable environments, production-scale performance | ||
| hardening, and installation from published Python packages. |
There was a problem hiding this comment.
Future releases will focus on:
- Enhanced newton support for benchmark involving contact-rich insertion, cables, and deformables.
- Prompt to scene+robot+task workflows
- Improved performance for multi-node evaluation.
| This release introduces prompt-first environment generation, typed environment and experiment | ||
| definitions, configurable evaluation variations and sensitivity analysis, expanded policy | ||
| integrations, and new Kitchen Benchmark and RoboLab task libraries. It also upgrades Arena's | ||
| Isaac Lab 3.0 interoperation and adds major improvements to object placement and validation. |
There was a problem hiding this comment.
This release introduces:
- prompt-first environment generation
- controlled randomization through variations
- sensitivity analysis
- multi-node evaluations
- expanded policy
- task libraries including RoboLab-style tasks and our own kitchenbench for tasks in realistic kitchens
Details below:
Signed-off-by: Qian Lin <qianl@nvidia.com>
Signed-off-by: Qian Lin <qianl@nvidia.com>
Signed-off-by: Qian Lin <qianl@nvidia.com>
Signed-off-by: Qian Lin <qianl@nvidia.com>
Signed-off-by: Qian Lin <qianl@nvidia.com>
- Replace the "What's Next" paragraph with the roadmap bullets from review: Newton support for contact-rich benchmarks, prompt-to-scene/robot/task workflows, and multi-node evaluation performance. - Convert the v0.3.0 lead paragraph into a bulleted summary of the release highlights, followed by the detailed sections. Signed-off-by: alex <amillane@nvidia.com>
73870b5 to
7a46cc3
Compare
|
|
||
| Isaac Lab-Arena is being developed as an open-source, shared evaluation framework that the community can | ||
| collectively enhance and expand. We invite you to try Isaac Lab-Arena 0.2 Alpha, share feedback, and help | ||
| collectively enhance and expand. We invite you to try Isaac Lab-Arena 0.3 Alpha, share feedback, and help |
There was a problem hiding this comment.
🟡 The Key Features list above is still the 0.2 one
This line now invites people to try 0.3 Alpha, but the Key Features block (lines 12-21) hasn't changed since the 0.2 notes — its four "New" markers are sequential task chaining, semantic placement, heterogeneous parallel evals, and RL/Lab interop. None of the 0.3 headliners you list just below at lines 69-75 show up there.
Since that block is the standing banner refreshed each release, could the "New" markers move onto the 0.3 items (prompt-first generation, variations + sensitivity analysis, multi-node evaluation, Kitchen Benchmark / RoboLab)?
| **Limitations** | ||
|
|
||
| - Installation from a published Python package is not yet supported; use the native ``uv`` source workflow or Docker. | ||
| - Performance is not yet hardened for production-scale workloads in Alpha stage. |
There was a problem hiding this comment.
🟡 Multi-run experiments fall back to CPU, and that isn't mentioned anywhere
disable_fabric_for_runs (isaaclab_arena/evaluation/run_execution.py:38 on release/0.3.0, from #1169) sets device="cpu" and disable_fabric=True for any experiment that builds the environment more than once — which is exactly the "environment rebuilds" and multi-node dispatch advertised at line 93. I couldn't find it documented in docs/ either, and #1169 isn't cited in the notes.
That's a real slowdown on a headline 0.3 workflow. Worth a Limitations bullet?
| - Performance is not yet hardened for production-scale workloads in Alpha stage. | |
| - Performance is not yet hardened for production-scale workloads in Alpha stage. | |
| - Experiments that build the environment more than once (multiple runs, or ``num_rebuilds > 1``) fall back to | |
| the CPU device with Fabric disabled, to work around a render-after-rebuild bug in Isaac Lab (#1169). |
## Summary Cherry-pick the v0.3.0 release notes onto the release branch. ## Detailed description - Cherry-picks 44d036c (#1228, merged to `main`) with no conflicts. - Adds the `v0.3.0` section to `docs/pages/references/release_notes.rst` and updates the intro to 0.3 Alpha, including the reviewed roadmap and release-highlight bullets. - Documentation only; `release_notes.rst` on this branch is now identical to `main`. Signed-off-by: Qian Lin <qianl@nvidia.com> Signed-off-by: alex <amillane@nvidia.com> Co-authored-by: qianlin <53278415+qianl-nv@users.noreply.github.com>
3 participants
Summary
Document the Isaac Lab-Arena v0.3.0 release.
Detailed description