Closed yang-yk closed 4 years ago
https://github.com/jshtok/RepMet/blob/22aec096cd89a06d008fdd5b335a816545bdf073/fpn/symbols/resnet_v1_101_fpn_dcn_rcnn_oneshot_v3.py#L953
Sorry to bother you. I wonder why there is no make_loss for RPN classification loss? @jshtok @duynn912
Hi @yang-yk, In MXNet the SoftmaxOutput is already a loss, so there is no need for MakeLoss.
Thanks!
https://github.com/jshtok/RepMet/blob/22aec096cd89a06d008fdd5b335a816545bdf073/fpn/symbols/resnet_v1_101_fpn_dcn_rcnn_oneshot_v3.py#L953
Sorry to bother you. I wonder why there is no make_loss for RPN classification loss? @jshtok @duynn912