fix: Jetson Orin install — pin numpy<2 first, drop [monolithic], gate lerobot behind Python 3.12+#138
Merged
rylinjames merged 2 commits intoMay 17, 2026
Conversation
…gpu resolution conflict
… lerobot behind Python 3.12+ Three bugs caused Jetson Orin Nano (JetPack 6 / Python 3.10) installs to fail: 1. numpy ABI crash: Jetson AI Lab torch and onnxruntime-gpu wheels are compiled against NumPy 1.x, but pip pulled numpy 2.x (no constraint). Fix: pin numpy<2 BEFORE installing torch/ort. 2. lerobot ResolutionImpossible: [monolithic] depends on lerobot==0.5.1 which requires Python >=3.12, but JetPack 6 ships Python 3.10. Fix: add python_version >= '3.12' marker to all lerobot pins; change Jetson extras from [serve,monolithic] to [serve] only. 3. torch from wrong index: install.sh pulled torch from standard PyPI (x86_64-only). Fix: pre-install torch from Jetson AI Lab index before pip install reflex-vla. Tested: bash -n install.sh, pytest tests/ (59 passed).
Collaborator
Author
|
LOOKS GOOD TO ME |
Collaborator
Author
|
@gemini-code-assist review |
rylinjames
added a commit
to DsThakurRawat/reflex-vla
that referenced
this pull request
May 19, 2026
The existing 10 m/s ceiling capped end-effector velocity, which made sense for arms but blocked legitimate aerial flight speeds. Quadcopter and delivery drones routinely operate at 15-50 m/s (54-180 km/h). Raises max from 10.0 -> 50.0 and polishes the description to make the dual-use semantics explicit. None of the four shipped presets are affected (franka 1.0, ur5 1.0, so100 0.5, quadcopter 5.0 — all well under both ceilings). Originally proposed by @DsThakurRawat in FastCrest#119; this commit applies just the actual schema delta on top of current main, since the PR branch predated FastCrest#132 (quadcopter preset) and FastCrest#134-FastCrest#138 — naively merging would have reverted that work. Co-Authored-By: DsThakurRawat <divyanshrawatofficial@gmail.com> Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
rylinjames
added a commit
that referenced
this pull request
May 19, 2026
…nts (#119) The existing 10 m/s ceiling capped end-effector velocity, which made sense for arms but blocked legitimate aerial flight speeds. Quadcopter and delivery drones routinely operate at 15-50 m/s (54-180 km/h). Raises max from 10.0 -> 50.0 and polishes the description to make the dual-use semantics explicit. None of the four shipped presets are affected (franka 1.0, ur5 1.0, so100 0.5, quadcopter 5.0 — all well under both ceilings). Originally proposed by @DsThakurRawat in #119; this commit applies just the actual schema delta on top of current main, since the PR branch predated #132 (quadcopter preset) and #134-#138 — naively merging would have reverted that work. Co-authored-by: RomirJ <playindus@gmail.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Fixes three bugs that cause
pip installto fail on Jetson Orin Nano (JetPack 6 / L4T R36.x / Python 3.10):torchandonnxruntime-gpuwheels are compiled against NumPy 1.x.install.shpinnednumpy<2after the mainpip installwhich already pulled numpy 2.x → both libraries crash on import.lerobotResolutionImpossible —[monolithic]extra pinslerobot==0.5.1which requires Python ≥3.12. JetPack 6 ships Python 3.10 → pip hard-fails.install.shpulled torch from standard PyPI (x86_64-only) instead of the Jetson AI Lab index.Changes:
pyproject.toml: Addedpython_version >= '3.12'markers to all 3lerobot==0.5.1pins ([native],[monolithic],[rtc]). Pip now skips lerobot on Python 3.10/3.11 instead of crashing.install.sh: Changed Jetson extras from[serve,monolithic]→[serve]. Moved GPU dep install (numpy<2, torch, ort) beforepip install reflex-vla. Added torch install from Jetson AI Lab index.examples/02-deploy-smolvla-jetson.md: Rewrote install section with correct order and two-step deploy workflow (export on desktop, serve on Jetson).docs/troubleshooting.md: Added two new entries under "Jetson-specific issues" for the NumPy ABI crash and lerobot resolution failure.Type of Change
How Has This Been Tested?
pytest tests/passes locally (59 passed)bash -n install.shsyntax check passesreflex doctorsanity checkChecklist