Issues with TensorRT on RTX 3090 - giexec: not found

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

Hello everyone.

I have been running benchmarks for ResNet for some time now, on a wide selection of GPU’s.
Now I am attempting to run the same benchmarks on an RTX 3090, but no luck with this so far.
The system I am running is using Ubuntu 20.4.
NVIDIA Driver version is 455.23.04
TensorRT 20.10-py3
TensorFlow 20.10-tf2-py3
CUDA cuda_11.1.1_455.32.00

A typical command line looks like this.

nvidia-docker run --shm-size=1g --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --rm -v ~/Downloads/models/:/models -w /opt/tensorrt/bin nvcr.io/nvidia/tensorrt:20.10-py3 giexec --deploy=/models/ResNet-50-deploy.prototxt --model=/models/ResNet-50-model.caffemodel --output=prob --batch=16 --iterations=500 --fp16

However, I get an error like this now.

/usr/local/bin/nvidia_entrypoint.sh: line 66 exec: giexec: not found

Ubuntu is not my usual working environment, I am a windows guy. So this error has left me confused as it worked just fine before.

Any idea on how this might be resolved ?

Thanks everyone !

2

Any ideas on giexec not found error ?

3

Hi @wharmon,
The sample utility giexec has been renamed to trtexec to be consistent with the product name since TRT 4. A compatibility script for users of giexec has been included to help users make the transition. Could you please try trtexec?
https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/trtexec
Thanks!

4

Thank you !

This works fine now :)

1 Like