AcceleratorError: CUDA error: no kernel image is available for execution on the device

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
,
1

Hi. I would appreciate any help to resolve the following issue while running transformers pipeline on my Jetson Orin Nano:

AcceleratorError: CUDA error: no kernel image is available for execution on the device

GPU Type: Jetson Orin Nano Super
CUDA Version: 12.6
Operating System + Version: Ubuntu 22.04.5 LTS
Python Version (if applicable): Python 3.10.12
TensorFlow Version (if applicable): 2.20.0
PyTorch Version (if applicable): 2.10.0+cu126
Transformers Version: 4.53.0

Thanks!

5

Hi,

The error indicates that the CUDA kernel was not built with the target device’s GPU architecture.
For Orin Nano, it should be sm_87.

Please try our PyTorch package below:

Thanks.

6

Hi. Thank you so much for your help. I appreciate it! I installed it and it worked. Thanks!

Just to fully document it for newbies: It also requires updating libcudss.so.0. This can be done as per one of your previous posts:

9

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.