Closed kbegiedza closed 2 years ago
AFAIK, tensorrt has bad int8 performance on layers like SiLU.
You can try old version like tensorrt7.0. Or try another model with only RELU as activation.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Please refer to our open source quantization tool ppq, the quantization result is better than the quantization tool that comes with tensorrt, almost the same as the float32 model. https://github.com/openppl-public/ppq/blob/master/md_doc/deploy_trt_by_api.md
Env
Ubuntu 20.04.3 LTS
Driver Version: 470.57.02
CUDA Version: 11.4
TensorRT: 7.2.2-1
About this repo
which branch/tag/commit are you using?
master
which model? yolov5, retinaface?
yolov5m
Your problem
I've created my
*.engine
from previously trained*.pt
. ForINT8
I've used ~10% of dataset to calibrate model.Model converted with
FP16
precision outputs correct classes with high accuracy, but model withINT8
had very poor accuracy (detections with probability <0.1
).