Failed to create engine from onnx file using 'buildEngine.py'

[user123137]

Hi
I failed to build the engine using the methods in this blog.
Speeding Up Deep Learning Inference Using TensorFlow, ONNX, and NVIDIA TensorRT | NVIDIA Technical Blog
The error message is as follows：
python3 buildEngine.py --onnx_file /home/nvidia/resnet50.onnx --plan_file resnet50.plan
Traceback (most recent call last):
File “buildEngine.py”, line 30, in
main(args)
File “buildEngine.py”, line 21, in main
engine = eng.build_engine(onnx_path, shape= shape)
File “/home/nvidia/下载/tf2_code/engine.py”, line 14, in build_engine
with trt.Builder(TRT_LOGGER) as builder,
AttributeError: enter

[SivaRamaKrishnaNV]

Dear @user123137,
May I know the Jetpack version? We have ResNet50.onnx model as part of TRT sample. Did you check using it directly?

[user123137]

Hi
Jetpack’s version is 4.5.1.Whether it is a low version？
I used Onnx.checker.check_model method to check it and it has no problem.

[SivaRamaKrishnaNV]

Dear @user123137,
Could you test with Jetpack 5.1

[user123137]

Thank you for your answer.
Can Jetpack 4.5.1 and L4T 32.5.1 be upgraded with package management tools instead of SD card and Sdkmanager?

[SivaRamaKrishnaNV]

Dear @user123137,
You should use sdkmanager to flash jetpack