About the parameters of CP-Cluster on COCO test-dev

[RangeKing]

I saw the examples about COCO val in README. Could you please provide the parameters(iou_threshold and min_score) of CP-Cluster on COCO test-dev? Thx!

[shenyi0220]

For all the repos with CP, my experience is just to keep the 2 params same between COCO-val and COCO-test-dev. Sure welcome to further discuss with me it they're not working for you.

[RangeKing]

Thanks for your reply. I followed your instruction and successfully compiled mmcv-full without any error or warning. But when I import mmcv.ops.

Do you know how to solve this problem? The environment information: PyTorch 1.10, python 3.9.12, CUDA 11.3.
I saw the same problem in #6 (comment) and #4 (comment).

[shenyi0220]

Ahh, not quite professional on this.
Let me share my env version with you: Python 3.8.12+Pytorch1.10.2+Cuda10.2

Also, let's try to compile standard MMCV without CP firstly. I'm not sure whether mmcv could support cuda 11+ well.

And from your log, looks your C++ part of mmcv is not compiled correctly, or something wrong with Cython library

[RangeKing]

Standard MMCV without CP can be successfully compiled and works well. I'll try to compile CP on CUDA 10.2.

[mastergao57]

Standard MMCV without CP can be successfully compiled and works well. I'll try to compile CP on CUDA 10.2.

Have you solved the problem undefined symbol: _Z3nmsN2at6TensorES0_fi?