train error, my cuda is 9.2

[tsing-cv]

Traceback (most recent call last):
  File "train.py", line 13, in <module>
    from libs.networks import build_whole_network
  File "../libs/networks/build_whole_network.py", line 13, in <module>
    from libs.box_utils import boxes_utils, iou_rotate
  File "../libs/box_utils/iou_rotate.py", line 10, in <module>
    from libs.box_utils.rbbox_overlaps import rbbx_overlaps
ImportError: ../libs/box_utils/rbbox_overlaps.cpython-36m-x86_64-linux-gnu.so: undefined symbol: __cudaPopCallConfiguration

@yangxue0827

[yangxue0827]

@tsing-cv

[tsing-cv]

compiled this codes as your discribled

[yangxue0827]

Delete the .so and .cpp files and recompile them. @tsing-cv

[tsing-cv]

I deleted the .so and .cpp files,but still can not solve this problem

[tsing-cv]

@yangxue0827 @yangJirui I used python36, cuda 9 encountered
no CUDA-capable device is detected,
when i used python27, cuda 8 encountered
undefined symbol: __cudaPopCallConfiguration
Do you have some suggests

My cuda version is cuda9.2, have no problem runing other projects

[yangxue0827]

Try to set value of GPU_GROUP in cfgs to choose your gpu index. Default is 2, you may dont have three gpus. @tsing-cv

[tsing-cv]

I seted GPU_GROUP="0";but still raise this
CUDA driver version is insufficient for CUDA runtime version
Other projects without these problems.
The devices is: gpu 1080ti, cuda9.2 nvidia-driver 384
@yangxue0827

[yangxue0827]

try CUDA_VISIBLE_DEVICES=0 python train.py @tsing-cv

[tsing-cv]

saved!!Thank you so much.