Trying to get onnxruntime_gpu on a Jetson agx orin Jetson 6 r36 and rev 4.4 Cuda 12.6

I have a Jetson AGX ORIN , R36, rev 4.4 cuda 12.6

I get a "“malloc(): invalid soze (unsorted)”

Aborted (core dump)

I have only 1 onnxruntime_gpu installed

Onnxruntime-gpu installation

Hi,

Please try the package below:

Thanks.

I am using jp6/cu126 and i get the

I get a "“malloc(): invalid soze (unsorted)”

Aborted (core dump)

Hi,

Could you share which command you used?
Do you meet the error when import the onnxruntime?

Thanks.

Hi, I have the same issue as mpatalberta.

I did pip install onnxruntime_gpu-1.23.0-cp310-cp310-linux_aarch64.whl, then python3, import onnxruntime which led to malloc(): invalid size (unsorted)
Aborted (core dumped).

Hi, both

We recently built and validated the onnxruntime_gpu with CUDA 12.8 below:

Could you give it a try?
Thanks.

Experiencing the same issue. Jetson Orin AGX, JetPack 6.2, CUDA12.6.
Tried both the cu126 and the cu128 versions of onnxruntime.
Both give the same malloc/free error upon import.
This is solved when changing to 50W but we need to stay at 30W power setting due to battery limitation. Please provide a solution for lower power settings.

That’s on 30W settings:

(hell) nvidia@ubuntu:/data/projects/signscene_ws$ pip list | grep onnx
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
onnxruntime-gpu 1.22.0
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
(hell) nvidia@ubuntu:/data/projects/signscene_ws$ python
Python 3.10.20 (main, Mar 11 2026, 17:41:27) [GCC 14.3.0] on linux
Type “help”, “copyright”, “credits” or “license” for more information.

import onnxruntime
exit()
free(): invalid pointer
Aborted (core dumped)

(hell) nvidia@ubuntu:/data/projects/signscene_ws$ pip uninstall onnxruntime-gpu
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
Found existing installation: onnxruntime-gpu 1.22.0
Uninstalling onnxruntime-gpu-1.22.0:
Would remove:
/data/miniconda3/envs/hell/bin/onnxruntime_test
/data/miniconda3/envs/hell/lib/python3.10/site-packages/onnxruntime/*
/data/miniconda3/envs/hell/lib/python3.10/site-packages/onnxruntime_gpu-1.22.0.dist-info/*
Proceed (Y/n)? Y
Successfully uninstalled onnxruntime-gpu-1.22.0
(hell) nvidia@ubuntu:/data/projects/signscene_ws$ pip install /data/misc/onnxruntime_gpu-1.23.0-cp310-cp310-linux_aarch64.whl
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
Processing /data/misc/onnxruntime_gpu-1.23.0-cp310-cp310-linux_aarch64.whl
Requirement already satisfied: protobuf in /home/nvidia/.local/lib/python3.10/site-packages (from onnxruntime-gpu==1.23.0) (7.34.0)
Requirement already satisfied: flatbuffers in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from onnxruntime-gpu==1.23.0) (25.12.19)
Requirement already satisfied: packaging in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from onnxruntime-gpu==1.23.0) (25.0)
Requirement already satisfied: sympy in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from onnxruntime-gpu==1.23.0) (1.14.0)
Requirement already satisfied: coloredlogs in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from onnxruntime-gpu==1.23.0) (15.0.1)
Requirement already satisfied: numpy>=1.21.6 in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from onnxruntime-gpu==1.23.0) (1.26.4)
Requirement already satisfied: humanfriendly>=9.1 in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from coloredlogs->onnxruntime-gpu==1.23.0) (10.0)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in /data/miniconda3/envs/hell/lib/python3.10/site-packages (from sympy->onnxruntime-gpu==1.23.0) (1.3.0)
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
Installing collected packages: onnxruntime-gpu
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
Successfully installed onnxruntime-gpu-1.23.0
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
(hell) nvidia@ubuntu:/data/projects/signscene_ws$ pip list | grep onnx
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
onnxruntime-gpu 1.23.0
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
WARNING: Ignoring invalid distribution -orch (/data/miniconda3/envs/hell/lib/python3.10/site-packages)
(hell) nvidia@ubuntu:/data/projects/signscene_ws$ python
Python 3.10.20 (main, Mar 11 2026, 17:41:27) [GCC 14.3.0] on linux
Type “help”, “copyright”, “credits” or “license” for more information.
import onnxruntime
exit()
free(): invalid pointer
Aborted (core dumped)

I built a docker and that solved my problem.
I can send the dockerfile and support files… Who were to?

Patrick Martin Brown
Home: 631 775 7907
Work Cell : 631 880 1188
pmh786@zebra.com
mpatalberta@yahoo.com
mpatalberta@gmail.com

Hi,

We have verified that the cu128 package can work.
Please note that you will need to upgrade your CUDA to 12.8 as well:

Thanks.