kendryte / nncase

Open deep learning compiler stack for Kendryte AI accelerators ✨
Apache License 2.0
752 stars 183 forks source link

yolox-nano 在k210量化损失 #1042

Open LYHTZHANGJIANG opened 1 year ago

LYHTZHANGJIANG commented 1 year ago

Describe the bug A clear and concise description of what the bug is.

To Reproduce Command line or scripts to reproduce the behavior:

Expected behavior A clear and concise description of what you expected to happen.

Origin model and code If applicable, add orgin model and code to help explain your problem.

Environment (please complete the following information):

Additional context Add any other context about the problem here.

LYHTZHANGJIANG commented 1 year ago

nncase1.6.0,就跑的官方的20个类别的demo。 1、编译cpu的定点模型,输出是正常。 2、但是编译210的定点模型 python tools/compile.py model/yolox_nano_224.onnx yolox_nano_224_quant.kmodel --imgs_dir ../20classes_yolo/images/ --legacy --target k210 python tools/simulate.py yolox_nano_224_quant.kmodel ../20classes_yolo/images/dog.bmp 这张dog.bmp的图片只能输出一个车,并且还没框准确。 res_cpu_quant ![Uploading res_K210_quant.jpg…]()