ONNX -> TRT Error Code 1 and 2: Cask (isConsistent) and Internal Error (Assertion enginePtr != nullptr failed.)

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Description

I’m trying to deploy ONNX model using TensorRT, but errors occours.
image

Used command:
./trtexec --onnx=/home/nvidia/trt_conversion/build-yolo-tensorrt/model/tfmodel_lin_19_batchsize.onnx --saveEngine=test.trt --verbose

The model was converted to ONNX from ‘SavedModel’ format on target Device and check with onnx.checker.check_model(model). do not return any error.

I tried conversion with and without set an explicit batch size in the ONNX file.

The attempt was made also with onnx==1.4.1.
The attempts were also made when convert model on host machine with different opset (9, 11, 12).

Environment

JP Version: 4.6.0
TensorRT Version: 8.0.1.6
Device Type: Jetson Xavier
Nvidia Driver Version:
CUDA Version: 10.2.460-1
Python Version (if applicable): 3.6.9
Tensorflow Version (if applicable):tensorflow==2.6.2+nv21.12

Relevant Files

Link to ONNX model (without set an explicit batch size- the error is the same for both cases) - Sign in to your account

Steps To Reproduce

Please include:
Commands in description.

2

Hi,
Request you to share the ONNX model and the script if not shared already so that we can assist you better.
Alongside you can try few things:

  1. validating your model with the below snippet

check_model.py

import sys
import onnx
filename = yourONNXmodel
model = onnx.load(filename)
onnx.checker.check_model(model).
2) Try running your model with trtexec command.

In case you are still facing issue, request you to share the trtexec “”–verbose"" log for further debugging
Thanks!

4

Added putty.log with --verbose: putty.log (305.8 KB)
Check_model don’t return anny errors.
Link to ONNX file (the same as in description)

5

Hi,

We could not reproduce the same error on the latest TensorRT version 8.4 GA.
We recommend you to please use the latest TensorRT version.

&&&& PASSED TensorRT.trtexec [TensorRT v8401] # /opt/tensorrt/bin/trtexec --onnx=tfmodel_lin_19.onnx --verbose

As you’re using Jetson Xavier, we recommend you to please reach out Jetson Xavier forum if you face any issues in upgrading the TRT version.

Thank you.

6

Hi,

Thanks for the advice, before upgrade I tried with adding input shape flag and explicit batch --shapes=input_1:1x416x416x3 --explicitBatch which solved the issue.

7

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