How to due with the undefined op in tensorflow

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

Will the tensorrt skip undefined op automatically in tensorflow ? Is it the only way to use the IPlugin to implementation undefined op? When the tensorrt meet the undefined op,it report a message like this :[TensorRT] ERROR: UFFParser: Validator error: conv_conv2_dilate/convolution/required_space_to_batch_paddings/mod_1: Unsupported operation _FloorMod.

CUDA version:10.0
CUDNN version:7.3.1
Python version [if using python]:python3
Tensorflow version:1.13rc0
TensorRT version :TensorRT-5.0.2.6.Ubuntu-16.04.4.x86_64-gnu.cuda-10.0.cudnn7.3

2

In order to use UFF, all the ops have to be known to UFF and TensorRT.
Instead you can use TF-TRT which leaves the unknown ops in TF, and runs the known ops in TensorRT.

Examples: GitHub - tensorflow/tensorrt: TensorFlow/TensorRT integration

Presentation: https://developer.nvidia.com/gtc/2019/video/S9431

User guide: https://docs.nvidia.com/deeplearning/dgx/tf-trt-user-guide/index.html