zybsu27f22 opened a new issue #18966: URL: https://github.com/apache/incubator-mxnet/issues/18966
## Description gluon nd接口与sym接口运算结果不一致的问题：如果在使用gluon定义的网络 中输出的结果有加上常数的时候模型hybridize后计算结果会出现错误。 ## To Reproduce 比如在gluoncv 中yolov3模型将YOLOOutputV3中box_centers = F.broadcast_add(F.sigmoid(raw_box_centers), offsets) * self._stride 改为 box_centers = F.broadcast_add(F.tanh(raw_box_centers)*3 +1 , offsets) * self._stride，模型在hybridize后与nd计算结果不同。如果单独测试tanh(x) + 1 sym接口与nd计算结果一致，该问题只在定义好的gluon多层网络模型中存在单层网络中没法复现。 ## What have you tried to solve it? 问题可能在模型hybridize过程中 ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
