ETLT VS TRT export and where to dump it in DS5

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Using Xavier AGX, DS5, TRT 7.
We trained a yolo based model using TLT2 and want to deploy on Xavier AGX, yet we get “Segmentation Error” when launching DS5.
I suspect it is due to a TRT export that used a V100 now being deployed on Jetson AGX Xavier.
When exporting from TLT2, do you export an ETLT or TRT?
Once ETLT is exported, in which directory do you dump the engine, weights, config and labels file in DS5?
We built a single class using YoloV3 so I assume my label.txt only needs one class right?
TLT2, requires only 2 lines of code…just like a 747 who requires two clicks to go on autopilot.

• Hardware Platform (Jetson / GPU)
• DeepStream Version
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)

3

Hi,

You will need to use ETLT format.
TensorRT engine is not portable so cannot be used cross software or platform.

Here are some TLT for Deepstream sample for your reference:

Thanks.