Add jetson-trt reference (TensorRT + GPU, native and in-container on Jetson)#24
Open
mobileoverlord wants to merge 1 commit into
Open
Add jetson-trt reference (TensorRT + GPU, native and in-container on Jetson)#24mobileoverlord wants to merge 1 commit into
mobileoverlord wants to merge 1 commit into
Conversation
…Jetson)
Demonstrates the integrated Tegra GPU on JetPack 7 / L4T r39.2 targets two
ways, selectable at provision time:
-r native : host-native TensorRT inference (python3-tensorrt from the feed,
no container engine)
-r docker : GPU inside a container via upstream docker (moby) + the NVIDIA
Container Toolkit's runtime (Tegra CSV passthrough),
running a pre-seeded CUDA image on the iGPU
Default target jetson-agx-thor (T264); also supports jetson-agx-orin and
jetson-orin-nano, which share the 2026 feed. Uses the current permissions:
profile syntax (dev profile baked into rootfs + initramfs).
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
A new reference, jetson-trt, demonstrating the integrated NVIDIA Tegra GPU on JetPack 7 / L4T r39.2 Jetson targets running TensorRT — both host-native and inside a container — modeled on the
x86-rtxreference but done the Jetson way.Two runtimes, selectable at provision time:
-r nativepython3-tensorrt+ CUDA runtime from the feed, no container engine-r dockernvidiaruntime (Tegra CSV passthrough), running a pre-seeded CUDA image on the iGPUNotes
nvidia-container-toolkit. Unlikex86-rtx(CDI,--device nvidia.com/gpu=all), Tegra uses the toolkit'snvidiaruntime in CSV-passthrough mode — the host GPU userspace (incl.nvidia-smi,libcuda) is bind-mounted in fromtegra-configs-container-csv, andnvidia-container-setup.servicewrites the runtime config at boot.permissions:profile syntax (adevprofile baked into both rootfs and initramfs, so the emergency console is reachable when debugging a failed boot).nvcr.io/nvidia/cuda:13.2.1-runtime-ubuntu24.04(matches Thor's CUDA 13.2; arm64) — NVIDIA has not published the classicl4t-*base images for JP7.Targets
jetson-agx-thor(T264) — tested first, live hardware bring-up.jetson-agx-orin,jetson-orin-nano(share the 2026 feed).Status
Native + docker runtimes validated on Thor hardware:
nvidia-smiand a CUDA workload run on the iGPU inside the container via--runtime nvidia. Depends on the 2026 feed carrying the Jetson packages (in progress on the build cluster).