TensorRT model consuming more amount of RAM

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hi Team,
We are using TensorRT .pb model on the Jetson TX2 platform, and the model size is 165MB. Once the model gets loaded, it consuming more amount of RAM ( RAM usage is increased around 1.5GB to 5.3GB, CPU usage is 20% and Maximum GPU usage is 39 %). Is it normal behavior? or we can optimize the RAM with some system configurations?. Kindly help us on this issue.

Below I am mentioning package versions which we used
Jetpack - 4.4
CUDA - 10.2
Tensorflow - 2.2.0
Libcudnn - 8.0
Linux ubuntu - 18.04
TensorRT - 7.1.3

2

Hi @vignesh.nanjundan,
Jetson team should be able to help you better here, hence moving this to Jetson Team.
Thanks!

4

Hi,

You can limit the TensorRT memory usage by setting workspace variable.
Please check this comment for more information:

Thanks.