Torch CUDA DSA question

[justmore5mins]

I uses Jetson Orin Nano on Jetpack 6.2, when I executed yolo export model=“best.pt” data=“path/to/dataset” format=“engine” and it returns back
Ultralytics 8.3.75 🚀 Python-3.10.12 torch-2.6.0+cu126 CUDA:0 (Orin, 7620MiB)
Traceback (most recent call last):
File “/home/jetson/Desktop/venv/bin/yolo”, line 8, in
sys.exit(entrypoint())
File “/home/jetson/Desktop/venv/lib/python3.10/site-packages/ultralytics/cfg/init.py”, line 986, in entrypoint
getattr(model, mode)(**overrides) # default args from model
File “/home/jetson/Desktop/venv/lib/python3.10/site-packages/ultralytics/engine/model.py”, line 741, in export
return Exporter(overrides=args, _callbacks=self.callbacks)(model=self.model)
File “/home/jetson/Desktop/venv/lib/python3.10/site-packages/ultralytics/engine/exporter.py”, line 330, in call
model = model.fuse()
File “/home/jetson/Desktop/venv/lib/python3.10/site-packages/ultralytics/nn/tasks.py”, line 201, in fuse
m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv
File “/home/jetson/Desktop/venv/lib/python3.10/site-packages/ultralytics/utils/torch_utils.py”, line 257, in fuse_conv_and_bn
w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var)))
RuntimeError: CUDA error: no kernel image is available for execution on the device
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.
my cuda version is 12.6 which same as torch cuda version, can anyone help me solve the problem?

[AastaLLL]

Hi,

Could you share which PyTorch package do you install?
Thanks.

[justmore5mins]

this version
I install it by pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

[AastaLLL]

Hi,

Could you try our prebuilt to see if it works?

https://pypi.jetson-ai-lab.dev/jp6/cu126

Thanks.

[justmore5mins]

Because there might be a lot of problem i need to solve, so i downgrade it to jp6.1, hope it will execute successfully

[AastaLLL]

Hi,

It’s recommended to install our PyTorch as the package from PyTorch’s official website might not be built with Jetson forum support.

YOu can find the step-by-step tutorial for YOLO in the below link:
https://elinux.org/Jetson/L4T/TRT_Customized_Example#YoloV11_using_Pytorch

Thanks.

[justmore5mins]

thank you very much!