Onnx to tensorrt conversion fails

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

Description

trtexec --onnx=my_model.onnx --batch=1 --saveEngine=test.engine --verbose
fails with the below error
[06/16/2021-12:03:24] [V] [TRT] ImporterContext.hpp:116: Registering tensor: 498 for ONNX tensor: 498
[06/16/2021-12:03:24] [V] [TRT] ModelImporter.cpp:179: Cast_9 [Cast] outputs: [498 → ()],
[06/16/2021-12:03:24] [V] [TRT] ModelImporter.cpp:103: Parsing node: Neg_10 [Neg]
[06/16/2021-12:03:24] [V] [TRT] ModelImporter.cpp:119: Searching for input: 498
[06/16/2021-12:03:24] [V] [TRT] ModelImporter.cpp:125: Neg_10 [Neg] inputs: [498 → ()],
ERROR: onnx2trt_utils.cpp:1686 In function unaryHelper:
[8] Assertion failed: validUnaryType
[06/16/2021-12:03:24] [E] Failed to parse onnx file
[06/16/2021-12:03:24] [E] Parsing model failed
[06/16/2021-12:03:24] [E] Engine creation failed
[06/16/2021-12:03:24] [E] Engine set up failed

Environment

TensorRT Version: 7.1.3.4
GPU Type: GeForce RTX 3090 , 1070
Nvidia Driver Version: 460.32.03
CUDA Version: 11.2
CUDNN Version: 8.1.1.33
Operating System + Version: ubuntu 20.04

Python Version (if applicable):
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag):

Relevant Files

modelmodel
loglog

Steps To Reproduce

trtexec --onnx=midas.onnx --batch=1 --saveEngine=test.engine --verbose

Please include:

  • Exact steps/commands to build your repro
  • Exact steps/commands to run your repro
  • Full traceback of errors encountered
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.
https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/trtexec
In case you are still facing issue, request you to share the trtexec “”–verbose"" log for further debugging
Thanks!

5

thanks for your response ! I did share the links of log, the model earlier. Please find them my_model.onnx - Google Drive
trt.log - Google Drive
Inference from the onnxmodel works fine.
The following snippet outputs
input.1
input shape [1, 3, 512, 512]
output name 1349

import onnxruntime
model_name = “my_model.onnx”
sess = onnxruntime.InferenceSession(model_name)

input_name = sess.get_inputs()[0].name
print(input_name) # input.1

input_shape = sess.get_inputs()[0].shape
print(“input shape”, input_shape) # [1, 3, 384, 672]

output_name = sess.get_outputs()[0].name
print(“output name”, output_name)

6

Hi @eyebies,

It is known issue. Fix will be available soon in future releases.

Thank you.

1 Like
7

Thanks polisetty for your reponse ! Would it be possible to explain whats the actual cause so that I could alter my model !
thanks again!

8

@eyebies,

The main reason for this issue is TRT currently does not support INT32 types for the NEG operator. Fix for this will be available soon in future releases.

1 Like
9

Hello. Any updates regarding the issue? Currently using TRT v. 7.2.1 and got the same error.