Help? Training my own dataset is done, test on images error : Check failed: error == cudaSuccess (77 vs. 0) an illegal memory access was encountered *** Check failure stack trace: ***

[brisker]

I1020 12:30:46.939435 28978 net.cpp:865] Ignoring source layer rpn_loss_bbox I1020 12:30:46.939491 28978 net.cpp:865] Ignoring source layer rpn_loss_cls I1020 12:30:46.939963 28978 net.cpp:865] Ignoring source layer seg_cls_score_ext_seg_cls_score_ext_0_split I1020 12:30:46.939996 28978 net.cpp:865] Ignoring source layer seg_cls_score_seg_cls_score_0_split F1020 12:30:47.198374 28978 syncedmem.hpp:18] Check failed: error == cudaSuccess (77 vs. 0) an illegal memory access was encountered * Check failure stack trace: *

[hgaiser]

For future reference, I'm assuming you ran out of memory. Are you running VGG16? I believe you require at least 6GB of memory to train VGG16.