TensorRT conversion fails with DCHECK(!i->is_use_only())

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

Description

I am trying to convert the following simple ONNX model to TensorRT:
https://cloud.safead.de/s/Wq2xm3dgZLoHXmx

Both when using trtexec or the python API it fails with:
operation.cpp:203: DCHECK(!i->is_use_only()) failed.

Environment

TensorRT Version: 8.5.2.2
GPU Type: A6000
Nvidia Driver Version: 525.60.13
CUDA Version: 11.8
CUDNN Version: 8.7.0
Operating System + Version: Ubuntu 22.04
Python Version (if applicable): 3.10.6
PyTorch Version (if applicable): 1.12.1
Baremetal or Container (if container which image + tag): Baremetal

Relevant Files

Please attach or include links to any models, data, files, or scripts necessary to reproduce your issue. (Github repo, Google Drive, Dropbox, etc.)

Steps To Reproduce

Run trtexec --onnx=test.simple.onnx --saveEngine=test.trt

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

You can find the ONNX model in the link in my first post.

onnx.checker.check_model does not report any error for the ONNX model.

The verbose output of trtexec can be found here: SafeAD Cloud

5

Hi,

We could reproduce the same error. Please allow us sometime to work on this.

Thank you.

1 Like
6

Hello, I have met the same error, is there any progress?

7

The input size of matmul is so large to cause this bug

8

Any improvement on this topic? I have the same error during Trtexec conversion.

9

In my case there was a torch.matmul usage with dimension broadcasting (other than batch dim). It caused this problem. When I removed the broadcast and handled it with transposes etc. it is resolved.

2 Likes
10

Thank you very much! I have solved this error by modifying the code related to “torch.matmul”.
For example:
Before:
C = A.view(b,m,n,c,h,w).matmul(B)
After modification:
C = A.view(b,m,n,c,h,w)
C = C.matmul(B)

11

have you fix it?