Segmentation Fault 11 when using mx.ndarray.argmax_channel

[guoqiang0148666]

Thank your excellent works.

l train cascade_rcnn for two-class objection detection, after some epochs, l meet Segmentation Fault 11 when calculating 'RcnnAcc_1st', l debug the code and think the error may exist in the bellow code

In core/detection_metric.py

pred_label = mx.ndarray.argmax_channel(pred).astype('int32').asnumpy().reshape(-1)

l just wonder if it right and how to solver it.

Any suggestions will be welcomed.

Thanks