search

Deci-AI / super-gradients

Easily train or fine-tune SOTA computer vision models with one open source training library. The home of Yolo-NAS.
https://www.supergradients.com
Apache License 2.0
4.59k stars 509 forks source link

Unable to detect target after qat #1147

Open yyj-1029 opened 1 year ago

yyj-1029 commented 1 year ago

I trained my model on my own dataset following the guidance from qat_ptq_yolo_nas.md. The model works fine in non-QA situations, but after applying QAT, it fails to detect any objects.

ofrimasad commented 1 year ago

Hi @yyj-1029 . When you say it fails to detect any objects, do you refer to the ONNX file, the TRT engine or the original torch model? Can you share some code? Please note that when performing QAT/PTQ the model goes through some changes that prepare it for 8bit TRT. at that stage, it is saved to an ONNX file.