Trtexec and dynamic input shapes

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

Description

I am trying to convert a Tensorflow model to TensorRT.
My model takes one input: ‘input:0’ and outputs a ‘Identity:0’.
I want to use trtexec to generate an optimized engine for dynamic input shapes, but It’s been blocking at the phase of autotuning format combination for hours.

figure as follow:

xxxx
xxxx1239×647 64.2 KB

shapes as follow：
shapes
But when i use fixed input shapes，It’s working fine.

Any guidance here would be appreciated.

Environment

TensorRT Version: TensorRT-7.0.0.11.Windows10.x86_64.cuda-10.2.cudnn7.6
GPU Type: NVIDIA RTX2060
Nvidia Driver Version: 441.22
CUDA Version: 10.2
CUDNN Version: 7.6.5
Operating System + Version: windows10
TensorFlow Version (if applicable): tensorflow-gpu 2.3.0

2

Hi @chenhrf915,
Can you pls share your model so that we can assist you better.
Thanks!

6

I just found out that your official website has released 7.2.1 package to support windows. run without problem based on that version.
Anyway thank you.