Pre-trained model conversion to tensorrt format from etlt using tlt-converter on Orin NX

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi

I want to convert this (yolov4_tiny_usa_deployable.etlt) model to tensorrt format to run on Orin NX.

From what I understand one way to do it is using use tlt-converter. However my Orin NX runs Jetpack (5.1.1-b56) with cuda (11.4.19-1) and tensorrt (8.5.2.2-1+cuda11.4) and tlt-convert doesn’t support this version (as seen here) on Orin NX.

Can someone suggest how to convert the pre trained etlt yolov4 model to tensorrt format in Orin NX?

Thanks!

5

You can refer to TAO Converter | NVIDIA NGC and download v4.0.0_trt8.5.2.2_aarch64 version for your case.

wget --content-disposition 'https://api.ngc.nvidia.com/v2/resources/org/nvidia/team/tao/tao-converter/v4.0.0_trt8.5.2.2_aarch64/files?redirect=true&path=tao-converter' -O tao-converter

You can also decode the .etlt model to .onnx file(refer to Fpenet retraining output file onnx but deepstream is using tlt - #12 by Morganh) and then use trtexec to generate tensorrt engine. Refer to TRTEXEC with YOLO_v4_tiny - NVIDIA Docs.

6

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.