Python deepstream app 5.0 generate different results every time I rebuild the engine

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

Hello,

I’m using python version 1.0 of deepstream 5.0 with yolov3-tiny. However, my deepstream app was implemented on python version 0.9 and deepstream 5.0 DP and it was working perfectly and achieved good results. After the new version, I get different results every time I rebuild the engine, also each engine has different size. I rebuilt the model multiple times on the Jetson AGX and my workstation, and it changes each time.

Any ideas on what might cause the problem?