fix cuda compile error

[SFMDI]

to eliminate cuda compilation errors in the latest version of Visual Studio 16.9

Code modified to address the following issues
Fixes #888

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Merging #891 (d0cb25e) into master (03491ed) will increase coverage by 0.80%.
The diff coverage is 66.94%.

❗ Current head d0cb25e differs from pull request most recent head 2c496b6. Consider uploading reports for the commit 2c496b6 to get more accurate results

@@            Coverage Diff             @@
##           master     #891      +/-   ##
==========================================
+ Coverage   64.86%   65.66%   +0.80%     
==========================================
  Files         148      149       +1     
  Lines        9189     9447     +258     
  Branches     1647     1718      +71     
==========================================
+ Hits         5960     6203     +243     
- Misses       2902     2924      +22     
+ Partials      327      320       -7     

Flag	Coverage Δ
unittests	65.66% <66.94%> (+0.80%)	⬆️

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

Impacted Files	Coverage Δ
mmcv/cnn/bricks/conv.py	100.00% <ø> (ø)
mmcv/engine/test.py	26.88% <26.88%> (ø)
mmcv/ops/deform_conv.py	61.76% <31.25%> (-4.36%)	⬇️
mmcv/runner/hooks/logger/wandb.py	68.75% <40.00%> (-6.25%)	⬇️
mmcv/runner/hooks/momentum_updater.py	60.60% <58.33%> (+2.27%)	⬆️
mmcv/runner/hooks/lr_updater.py	65.78% <62.22%> (+8.96%)	⬆️
mmcv/onnx/symbolic.py	40.60% <75.00%> (+0.71%)	⬆️
mmcv/runner/hooks/evaluation.py	87.24% <87.24%> (ø)
mmcv/cnn/__init__.py	100.00% <100.00%> (ø)
mmcv/cnn/builder.py	100.00% <100.00%> (ø)
... and 18 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 03491ed...2c496b6. Read the comment docs.

[wangruohui]

Basically, this MR modify all floor and ceil to floorf and ceilf to force to call the implementation for single precision, thus can avoid some problem on Windows. Based on comments from other users, this modification works and seems pass all tests.
But I am not sure whether this change affect behavior when meeting double or half.

[ZwwWayne]

Hi @SFMDI ,
Thanks for your kind contribution. Could you sing the CLA and fix the lint?

[ZwwWayne]

Basically, this MR modify all floor and ceil to floorf and ceilf to force to call the implementation for single precision, thus can avoid some problem on Windows. Based on comments from other users, this modification works and seems pass all tests.
But I am not sure whether this change affect behavior when meeting double or half.

Usually DCN with FP16 training is not tried. We may arrange some benchmarks on that if necessary.

[RunningLeon]

@grimoire Could you take a look at this PR and the issue?

[grimoire]

If you are worrying about half and double precision, could we cast the tensor dtype in deform_conv_cuda.cu DeformConvForwardCUDAKernelLauncher before feeding the tensor into the kernel?

[wangruohui]

@SFMDI
Hello, this MR is currently blocked because you haven't signed CLA . Would you please make it signed to unblock this MR? Thank you!

[SFMDI]

@wangruohui
I just signed CLA

[ZwwWayne]

pre-commit-config.yaml should be recovered. The lint issue should be fixed. For cpp code we use google style and it can be reformated in VSCode.

[OpenMMLab-Assistant-007]

Hi!
@SFMDI
First of all, we want to express our gratitude for your significant PR in the OpenMMLab project. Your contribution is highly appreciated, and we are grateful for your efforts in helping improve this open-source project during your personal time. We believe that many developers will benefit from your PR.

We would also like to invite you to join our Special Interest Group (SIG) private channel on Discord, where you can share your experiences, ideas, and build connections with like-minded peers. To join the SIG channel, simply message moderator— OpenMMLab on Discord or briefly share your open-source contributions in the #introductions channel and we will assist you. Look forward to seeing you there! Join us ：https://discord.gg/UjgXkPWNqA

If you have WeChat account，welcome to join our community on WeChat. You can add our assistant ：openmmlabwx. Please add "mmsig + Github ID" as a remark when adding friends：）
Thank you again for your contribution❤