Hi,
Can you share the code of TensorRT-INT8 in python. I have problem for calibration.
Thank you.
Hello there. I am new to nvidia, thanks for the information.
Hi, jalalmiry, johnmorganhere
We try this mode is only test FPS, so not need calibration.
Only execute command follow as:
/usr/src/tensorrt/bin/trtexec --onnx=./yolov3.onnx --workspace=26 --int8
/usr/src/tensorrt/bin/trtexec --onnx=./yolov3.onnx --workspace=26 --useDLACore=0 --int8
/usr/src/tensorrt/bin/trtexec --onnx=./yolov3.onnx --workspace=26 --useDLACore=1 --int8
If we have researched it, we will share it.
Hi, when i run trtexec with workspace size of 3GB (3000 MB) I still get a message saying
Some tactics do not have sufficient space to run. Increasing workspace size may increase performance.
Is there a config file that limits the space that TRT can use? How can i get rid of this limit?
What number of batches and streams are you using to get 100 FPS?
Hi MostafaTheReal,
Please open a new topic. Thanks