TensorRT4 ConvTranspose layer VERY SLOW inference speed.

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

cliu13 May 24, 2019, 6:23am 1

I’m experiencing slow inference speed for ConvTranspose Layer (deconv) in TensorRT4, as is shown in the jpg URL.

The network architecture is based on resnet-18 layers, with 32x downsampling and 8x upsampling(three deconv layers), where the three deconv layers take almost 1/2 time of the inference time. It doesn’t seem right when compared to Conv operations, as both layers are both convolution ops.

Topic		Replies	Views
cuda::DeconvolutionLayer works slowly TensorRT	1	577	May 27, 2019
The inference of [ Deconvolution + Other Operations ], for example [ Deconvolution + Convolution ] in tensorrt is slower than mxnet TensorRT	4	1115	May 18, 2020
The inference time of Deconvolution in tensorrt is slower than pytorch Triton Inference Server - archived tensorrt	0	789	April 15, 2020
TensorRT inference time extremely slow TensorRT	1	449	January 31, 2023
Deconvolution Layer runs super slow in TensorRT TensorRT	1	1603	August 2, 2018
TensorRT - Deconvolution layer slow inference DriveWorks	7	1892	October 9, 2018
Trtexec ConvTranspose cost to much time when inference TensorRT	1	418	August 9, 2022
Does TensorRT support conv3d with Tensor Core ? TensorRT	13	1868	April 26, 2021
TensorRT inference significantly slower using kernel size 11 vs 3 TensorRT tensorrt	3	521	February 10, 2021
Working with 1D in TensorRT TensorRT tensorrt	6	1664	September 15, 2022

TensorRT4 ConvTranspose layer VERY SLOW inference speed.

Related topics