zhangzhangzhang

Results 9 comments of zhangzhangzhang

I have the same problem.Do you solve it?

量化的模型是用官方给的en_PP-OCRv3_det_slim这个模型进行导出onnx模型

![1705885377621](https://github.com/daquexian/onnx-simplifier/assets/47705677/ea75ca53-a36f-4232-87a0-72850177cc27) This configuration can be used

我是用ch_ppocr_v3_det_converter.py 这个直接python ch_ppocr_v3_det_converter.py

paddle: backbone.conv.conv.weight ---- (8, 3, 3, 3) paddle: backbone.conv.bn.weight ---- (8,) paddle: backbone.conv.bn.bias ---- (8,) paddle: backbone.conv.bn._mean ---- (8,) paddle: backbone.conv.bn._variance ---- (8,) paddle: backbone.stage0.0.expand_conv.conv.weight ---- (8, 8, 1, 1)...

这个好像是在paddle模型的最后一层

[train.log](https://github.com/frotms/PaddleOCR2Pytorch/files/15451963/train.log) ch_PP-OCRv3_det_student.yml 这个config

> 这个好像是在paddle模型的最后一层 是第一层就出现了问题