Significant Increase in Initialization Time for YOLOX TensorRT Model on NVIDIA-Jetson Nano 4gb

[fyzanahammad]

After converting the YOLOX model to TensorRT, the initialization time on an NVIDIA Jetson Nano 4GB has unexpectedly increased from 1:30 minutes to approximately 5 minutes over a period of time. The pytorch model was converted into tensorRT using the trt module provided by yolox tools that outputs a model_trt.pth file. Is this issue related to caching of yolox?