using onnx parser

[yeuo67]

Hi there,

I’m trying to parse “onnx” model to “tensorrt” , the goal is finally to parse from tensorrt to h5 file,

for now, I don’t really know how to parse to tensorrt and I trying to look at tensorrt developer guide but it doesn’t really work, there is any good example out there, how to use onnx_parser?

TIA

YEHU

[NVES_R]

Hi,

For starters, you could see if trtexec works for you - trtexec is a convenient CLI tool that comes with TensorRT releases.

You could try something like:

trtexec --onnx=model.onnx

# See help command for a full list of flags
trtexec -h

For using the API directly to parse, you can checkout the TensorRT samples:

• The python samples currently come with the TensorRT release, in /opt/tensorrt/samples/python/introductory_parser_samples/onnx_resnet50.py
• A C++ sample can be found on the TensorRT GitHub here: https://github.com/NVIDIA/TensorRT/blob/master/samples/opensource/sampleOnnxMNIST/sampleOnnxMNIST.cpp