Person_reID_baseline_pytorch icon indicating copy to clipboard operation
Person_reID_baseline_pytorch copied to clipboard

Published 20 hours ago • verified publisher icon layumi

Optimizing the model on Edge Device

Open chetansarmak opened this issue 4 years ago • 3 comments

Can you help me in suggesting which is the best optimizing method to implement this in edge devices with GPU ? For e.g it runs @3 FPS on Jetson Nano , How can I increase the speed on it to 6 FPS?

chetansarmak avatar Oct 23 '20 19:10 chetansarmak

Hi @chetansarmak You may consider float16 or TensorRT.

layumi avatar Oct 25 '20 09:10 layumi

Oh great , Thank you @layumi I have tried out FP16, I wanted to try TensorRT , as that should increase the speed greater than 6 FPS I guess . have you tried or Is there anyone who tried it earlier , cause running this model consumes most of GPU utilization in tensorRT I feel, correct me if I am wrong ?

Thanks in advance

chetansarmak avatar Oct 25 '20 17:10 chetansarmak

@chetansarmak Are you able to run this on Jetson as tensorRT? I couldn't be able covert this model into a tensorRT inference. Can you share any codes or methods you have?

RisithPerera avatar Apr 15 '21 16:04 RisithPerera