On the Jetson Xavier NX development board, I tried converting my PyTorch ResNet-50 model to TensorRT and quantified using FP16 and INT8 respectively. But when I tried to inference in these modes, I found that the speed was the same, which puzzled me. Is this normal?

There is no update from you for a period, assuming this is not an issue any more. Hence we are closing this topic. If need further support, please open a new one. Thanks Hi, Just want to make sure first. Have you maximized the device performance with the following command? $ sudo nvpmodel -m 0…

Inference Speed

Robotics & Edge Computing Jetson & Embedded Systems Jetson Xavier NX

Topic		Replies	Views
Same inference speed with Resnet50 for int8 and fp16 Jetson Xavier NX jetson-inference	4	708	October 18, 2021
Int8 is not faster than fp16 on xavier Jetson AGX Xavier tensorrt	5	780	October 18, 2021
Same inference speed for INT8 and FP16 TensorRT	10	5934	October 12, 2021
After converting ssdMobilnet from the examples, the model is slower Jetson Xavier NX tensorrt	4	510	October 18, 2021
Human pose detection model (MoveNet) TensorRT conversion on NVIDIA Jetson Jetson Xavier NX tensorrt , tensorflow , jetson-inference	7	2723	June 16, 2022
Tensorrt can not speed up well TensorRT	7	1679	June 29, 2022
Inference Time is not stable TensorRT	10	1789	January 3, 2019
Lower performance with TRT than plain TF? Jetson Xavier NX tensorrt , jetson-inference	14	2030	October 18, 2021
How can we know we have convert the onnx to int8trt rather than Float32? TensorRT tensorrt	23	1921	June 14, 2021
Low Compute utilization of converted TensorFlow model during inference Jetson TX2	19	1749	October 18, 2021

Inference Speed

Related topics