[Fix] Fix compiling error on windows

[wangruohui]

Fix #169, related to #551 , but the author does not respond.

[CLAassistant]

All committers have signed the CLA.

[Wuziyi616]

@wangruohui Seems that we can successfully build these ops, but will cause errors when importing them. Can you successfully train models using these modified ops in your Linux server?

[codecov]

Codecov Report

Merging #766 (6e5725f) into master (78f4562) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #766   +/-   ##
=======================================
  Coverage   48.97%   48.97%           
=======================================
  Files         208      208           
  Lines       15852    15852           
  Branches     2532     2532           
=======================================
  Hits         7763     7763           
  Misses       7591     7591           
  Partials      498      498           

Flag	Coverage Δ
unittests	48.97% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 78f4562...6e5725f. Read the comment docs.

[Wuziyi616]

Hi @wangruohui, can you try re-training one model which uses these ops to benchmark the result? For example, PartA2Net that uses iou3d. If you have trouble downloading the data, I can train it for you.

[ZwwWayne]

Hi @wangruohui, can you try re-training one model which uses these ops to benchmark the result? For example, PartA2Net that uses iou3d. If you have trouble downloading the data, I can train it for you.

Please train one with this PR. Actually, if the related model can test well, we can merge that. The modification should be fine.

[Wuziyi616]

Have tested PointPillars on KITTI, PartA2 on KITTI, VoteNet on ScanNet and PAConv on S3DIS. They cover all the modified CUDA files in this PR. The test accuracy is good.

[wangruohui]

Thank you very much!