lerobot icon indicating copy to clipboard operation
lerobot copied to clipboard

support for running on Jetson?

Open liuhuanjim013 opened this issue 9 months ago • 1 comments

lerobot requires torchvision >0.21 which requires torch version >2.6 Latest nvidia torch version is 2.5 (for jetpack 6.1) and 2.4 (for jetpack 6.0)

I guess running on CPU is always an option, but the GPUs on Jetson is nice....

liuhuanjim013 avatar Mar 05 '25 10:03 liuhuanjim013

Got the following combo working for Jetson Orin NX (lerobot still requires torchvision 0.21 though)

liuhuanjim013 avatar Mar 09 '25 05:03 liuhuanjim013

Actually if you have Orin series you can install cuda 12.8 on you Jetson: https://developer.nvidia.com/cuda-12-8-0-download-archive?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local install cudnn 9.8.0 from here: https://developer.nvidia.com/cudnn-downloads?target_os=Linux&target_arch=aarch64-jetson&Compilation=Native&Distribution=Ubuntu&target_version=22.04&target_type=deb_local And use pytorch 2.7 from here: https://pypi.jetson-ai-lab.dev/jp6/cu128

Kracozebr avatar Apr 21 '25 06:04 Kracozebr

Tried the above but eventually ran into glibc issues.

JetPack 6.2 is based on Ubuntu 22.04 LTS (jammy), whose glibc is 2.35

The PyTorch wheel I got was built for manylinux 2_38 (≈ glibc 2.38, first shipped in Ubuntu 24.04) so it can’t run on jammy.

Glibc is the foundation of the whole OS; upgrading it in‑place is effectively an OS upgrade and isn’t supported on Jetson until NVIDIA publishes a new JetPack.

florinmatei avatar Apr 22 '25 03:04 florinmatei

@liuhuanjim013 were you able to run LeRobot on your Jetson ?

florinmatei avatar Apr 22 '25 03:04 florinmatei

Tried the above but eventually ran into glibc issues.

JetPack 6.2 is based on Ubuntu 22.04 LTS (jammy), whose glibc is 2.35

The PyTorch wheel I got was built for manylinux 2_38 (≈ glibc 2.38, first shipped in Ubuntu 24.04) so it can’t run on jammy.

Glibc is the foundation of the whole OS; upgrading it in‑place is effectively an OS upgrade and isn’t supported on Jetson until NVIDIA publishes a new JetPack.

What a shame. The next thing that you can do is to build pytorch 2.6.0 from source. I tried on my Jetson AGX Xavier but got errors due to old cuda version (11.4). With newer version I should build it without troubles.

Kracozebr avatar Apr 22 '25 17:04 Kracozebr

This issue has been automatically marked as stale because it has not had recent activity (6 months). It will be closed if no further activity occurs. Any change, comment or update to this issue will reset this count. Thank you for your contributions.

github-actions[bot] avatar Oct 20 '25 02:10 github-actions[bot]

This issue was closed because it has been stalled for 14 days with no activity. Feel free to reopen if is still relevant, or to ping a collaborator if you have any questions.

github-actions[bot] avatar Nov 03 '25 02:11 github-actions[bot]