Created by: bipedalBit

No matter in PC or PaddleCloud, when I change the CLASS_NUM for the SSD model other than 21（class num of VOC dataset) and use my own dataset, the training cannot run and always occur the similar error:

Thread [0x7fffb74d33c0] Forwarding detection_output,

* Aborted at 1522136207 (unix time) try "date -d @1522136207" if you are using GNU date *

PC: @ 0x0 (unknown)

* SIGFPE (@0x7fffae6f360a) received by PID 96282 (TID 0x7fffb74d33c0) stack trace: *

@ 0x7fffae7c6b3a _sigtramp Floating point exception: 8

Guess something wrong in paddle.layer.detection_output().