Created by: IvyGongoogle

Hello, when I use cascade_rcnn_mobilenetv3_fpn_640.yml to finetune my own dataset like:

python -u tools/train.py -c configs/mobile/cascade_rcnn_mobilenetv3_fpn_640.yml \
              -o pretrain_weights=https://paddlemodels.bj.bcebos.com/object_detection/cascade_rcnn_mobilenetv3_fpn_640.tar \
              finetune_exclude_pretrained_params=['cls_score','bbox_pred'] \
              --eval 

but after running about 2000 iters, it always show error:

Error Message Summary:
----------------------
Error: The start row index must be lesser than the end row index.
  [Hint: Expected begin_idx < end_idx, but received begin_idx:0 >= end_idx:0.] at (/paddle/paddle/fluid/framework/tensor.cc:83)
  [operator < multiclass_nms > error]

envs: python3.7, paddlepaddle 1.8.2.post97, paddleDetection-0.3

How to fix this? Can you give some advises? Thanks.