Skip to content

Latest commit

 

History

History

01-simple-convert-torch2trt

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 

trtexec-convert example

  • We are supposed that a GPU is used.
  • I used NGC Pytorch image to tensorrt converting.

1. Create trt model using ImageNet-pretrained resnet50

$ make convert-torch2trt
# You can get trt model in output dir
$ ls output
model.trt

2. trt model benchmarking

  • batchsize = 1
$ make benchmark
# Then, you can get performance summary
=== Performance summary ===
Throughput: 667.181 qps
Latency: min = 1.48242 ms, max = 1.59076 ms, mean = 1.49393 ms, median = 1.48682 ms, percentile(99%) = 1.57001 ms
End-to-End Host Latency: min = 1.49353 ms, max = 3.01935 ms, mean = 2.84119 ms, median = 2.85034 ms, percentile(99%) = 3.01239 ms

...

&&&& PASSED TensorRT.trtexec [TensorRT v8203] # trtexec --loadEngine=.//output/model.trt --batch=1

Reference