Efficientdet conversion from onnx to tensorrt engine freeze

Description

Running example from this Github

Freezing on the same spot in the last shown log. I am able to make it work using a workstation that using a quadro gpu.

Environment

TensorRT Version: 8202
GPU Type: RTX 3070
Nvidia Driver Version: 511.65
CUDA Version: 11.6
CUDNN Version: 8.3.2
Operating System + Version: ubuntu20.04 (windows 11 wsl)
Python Version (if applicable): 3.8
TensorFlow Version (if applicable): 2.7.0
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag): nvcr.io/nvidia/tensorflow:22.01-tf2-py3

Relevant Files

Log:
freeze.log (4.3 MB)

Onnx :
model.onnx (16.1 MB)

Steps To Reproduce

#!/bin/bash
INPUT_SHAPE=8x512x512x3
trtexec
–onnx=model/onnx-model/model.onnx
–saveEngine=model/trt-model/engine.trt
–optShapes=input:$INPUT_SHAPE
–workspace=8000
–verbose

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!

Hi,

We recommend you to please use the latest TensorRT version 8.4.
Using the latest TRT version, we could successfully generate the TRT engine.
https://developer.nvidia.com/nvidia-tensorrt-8x-download

Thank you.

Hi,

Ive tried the solution by using version 8.4, it still freeze. Tested by using windows zip installation on windows 11

update: after clear and kill some background process that using gpu resources, manage to make it work.

update2: another go with wsl, checkout latest, make using dockerfile provide by the repo instead of pulling from nvidia catalog, still freeze as the same spot as provided log

Hi,

We couldn’t reproduce the same error. We could successfully build the TRT engine.
Have you tried increasing workspace?
Could you please share with us Dockerfile to try from our end.

Thank you.

as mention, ive been using a container provide by nvidia but they didnt share the dockerfile, u need to pull this http://nvcr.io/nvidia/tensorflow:22.01-tf2-py3

but ive build another container suggest by official tensor rt repo and get the same freeze result, here is the shared dockerfile TensorRT/ubuntu-20.04.Dockerfile at main · NVIDIA/TensorRT · GitHub