szha edited a comment on issue #18800: URL: https://github.com/apache/incubator-mxnet/issues/18800#issuecomment-689127698
> > Any plan to fix MXNET_BACKWARD_DO_MIRROR, or to fix the docs since it is currently not supported by GLuon. I think MXNET_BACKWARD_DO_MIRROR is very important especially in object detection and sematic segmentation. Without that, even the backbone is resnet50, the im_per_gpu can only be set to 1 if the scale is 800 times 1333. > > @kohillyang or @szha do you want to open a PR with this fix for v1.8? I agree that this is an important feature though we may not make it in time for 1.8. ---------------------------------------------------------------- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
