Getting "Segmentation Fault" when converting RetinaNet model to TensorRT Engine File

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

Description

I am trying to convert RetinaNet ResNet 18 model available in the README of the follwing REPO:

However, I get an error of “Segmentation Fault” when converting the pth file to engine.plan file

  1. I downloaded the Resnet 18 and Resnet 34 pth models available in the README.md of this Repo GitHub - NVIDIA/retinanet-examples: Fast and accurate object detection with end-to-end GPU optimization
  2. I went through the list of commands to start the ODTK docker, namely:
git clone https://github.com/nvidia/retinanet-examples
docker build -t odtk:latest retinanet-examples/
docker run --gpus all --rm --ipc=host -it -v /home/vast/retinanet:/workspace/model odtk:latest
  1. Then ran the following command inside the docker in the /workspace/model directory:
odtk export retinanet_rn18fpn.pth engine.plan
  1. I get an error here:
Loading model from retinanet_rn18fpn.pth...
     model: RetinaNet
  backbone: ResNet18FPN
   classes: 80, anchors: 9
Exporting to ONNX...
Building FP16 core model...
Segmentation fault (core dumped)

How to debug this error?

Environment

TensorRT Version: 7.2.2
GPU Type: M60 Tesla
Nvidia Driver Version:
CUDA Version: 11.2
CUDNN Version:
Operating System + Version: Ubuntu
Python Version (if applicable): 3.6
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag): Built docker image using the following repo: GitHub - NVIDIA/retinanet-examples: Fast and accurate object detection with end-to-end GPU optimization

#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:
https://docs.nvidia.com/deeplearning/tensorrt/quick-start-guide/index.html#onnx-export

  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!

#5

I do not have an ONNX file. The command:

odtk export retinanet_rn18fpn.pth engine.plan

converts the .pth into the engine.plan file. Although it says:

Exporting to ONNX…

I do not know in which directory the ONNX file is being dumped (if at all the ONNX file is being created).

#6

Hi @subhankar.halder,

Please raise your concern on Issues · NVIDIA/retinanet-examples · GitHub to get better help.

Thank you.