TensorRT Benchmarking - How does it work without images?

[byyft2]

I’m measuring the performance of Jetson Nano

Deep Learning Inference Benchmarking Instructions

Hi all, below you will find the procedures to run the Jetson Nano deep learning inferencing benchmarks from this blog post with TensorRT. note: for updated JetPack 4.4 benchmarks, please use github.com/NVIDIA-AI-IOT/jetson_benchmarks [image] While using one of the recommended power supplies, make sure you Nano is in 10W performance mode (which is the default mode): $ sudo nvpmodel -m 0 $ sudo jetson_clocks Using other lower-capacity power supplies may lead to system instabilities or shutd…

I am proceeding by referring to the site above.

ex) ./trtexec --output=prob --deploy=…/data/googlenet/ResNet50_224x224.prototxt --fp16 --batch=1

I understand that if don’t set weights when trtexec proceed with benchmarks, trtexec will benchmark them with random weights.

But I didn’t set an input image for the above command.
How does it benchmark?
Does this situation also use random tensor?

I’d appreciate your reply.

[AakankshaS]

Hi @byyft2

@byyft2

But I didn’t set an input image for the above command.
How does it benchmark?
Does this situation also use random tensor?

The trtexec tool has two main purposes:

• It’s useful for benchmarking networks on random data.
• It’s useful for generating serialized engines from models.

Please check the below link for details
https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/trtexec

Thanks!

[byyft2]

@AakankshaS
Thank you very much for your reply.