mmdeploy
mmdeploy copied to clipboard
AssertionError: TensorRT plugin should be compiled
Hey I am developing mmocr on nvidia jetson nano. When I try to onnx2tensorrt deployment with the psnet_ctw detection model, I take this error. I read the documents about this issue but I couldn't find anything.
Please python3 tools/check_env.py
and find tensorrt item.
If it shows none
, follow this tutorial and install pytensorrt and pycuda.
Hi, @Paschapy Deployment tools in mmocr are already deprecated now. We encourage users to use mmdeploy instead.
Hi, @Paschapy Deployment tools in mmocr are already deprecated now. We encourage users to use mmdeploy instead.
Okey I got it. But I couldn't installation on jetson nano, when I tried to install mmdeploy I took an error. I looking for solution only I think it is unsolutions. My currently problem is python version (3.6.9) in Jetpack SDK. I tried to uninstall python3.6.9 and my jetson broken, I tried archiconda and I take pip3 error.
Please
python3 tools/check_env.py
and find tensorrt item.If it shows
none
, follow this tutorial and install pytensorrt and pycuda.
Hi I made progress, I finished installation. And I ran tools/deploy.py I took a different error. Like that;
I have no idea what to do for solution.
@AllentDan could you help investigate it?
You have to put icdar2015 dataset into a relative folder data .
Thank you for recommendation, I did it but took a different error like this,
Traceback (most recent call last):
File "tools/deploy.py", line 12, in
When I made a progress, I'm getting back. I formatted the jetson nano, and I followed the installation docs. But it's not worked. Maybe problem is my OS so Jetpack SDK 4.6.1 Jetson Nano version. I don't know.
From the picture you gave, it is the file not found error.
@AllentDan thank you for helping. You can close the issue.