add comments for ops LastLevelMaxPool to avoid potential confusion

[kkaiwwana]

This PR added some comments for ops LastLevelMaxPool which not really applied a max_pool2d to input but did subsampling actually. These comments attempt to clarify that and avoid potential confusion in future.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/7593

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 22 New Failures

As of commit a49b786:

NEW FAILURES - The following jobs have failed:

• build / linux-job (gh)
• c-source / linux-job (gh)
• onnx / linux-job (gh)
• python-source-and-configs / linux-job (gh)
• python-types / linux-job (gh)
• unittests-extended / linux-job (gh)
• unittests-linux (3.10, linux.12xlarge, cpu) / linux-job (gh)
• unittests-linux (3.11, linux.12xlarge, cpu) / linux-job (gh)
• unittests-linux (3.8, linux.12xlarge, cpu) / linux-job (gh)
• unittests-linux (3.8, linux.g5.4xlarge.nvidia.gpu, cuda, 11.7) / linux-job (gh)
• unittests-linux (3.9, linux.12xlarge, cpu) / linux-job (gh)
• unittests-macos (3.10, macos-12) / macos-job (gh)
• unittests-macos (3.11, macos-12) / macos-job (gh)
• unittests-macos (3.8, macos-12) / macos-job (gh)
• unittests-macos (3.9, macos-12) / macos-job (gh)
• unittests-prototype (3.8, linux.12xlarge, cpu) / linux-job (gh)
• unittests-prototype (3.8, linux.g5.4xlarge.nvidia.gpu, cuda, 11.7) / linux-job (gh)
• unittests-prototype (3.9, linux.12xlarge, cpu) / linux-job (gh)
• unittests-windows (3.10, windows.4xlarge, cpu) / windows-job (gh)
• unittests-windows (3.11, windows.4xlarge, cpu) / windows-job (gh)
• unittests-windows (3.8, windows.4xlarge, cpu) / windows-job (gh)
• unittests-windows (3.9, windows.4xlarge, cpu) / windows-job (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[facebook-github-bot]

Hi @iaoqian!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[vfdev-5]

Thanks for the PR @iaoqian , however, checking Detectron2 code, they also have kernel_size=1:
https://github.com/facebookresearch/detectron2/blob/d4a5f28e01b2babbaba9f90198fb95f5c661cccd/detectron2/modeling/backbone/fpn.py#L200

and same for Detectron implementation:
https://github.com/facebookresearch/Detectron/blob/6020e9a5aaaa21316cb87a8bc89605e2594310e7/detectron/modeling/FPN.py#L229

Seems like it is intended to have kernel_size=1 as they left a comment:

# Use max pooling to simulate stride 2 subsampling

[NicolasHug]

Also from the paper:

[kkaiwwana]

Well, Thanks @vfdev-5 @NicolasHug for your reply. That makes sense. In this way, LastLevelMaxPool may not be a proper name, I think. Maybe we can refactor this class with name 'LastLevelSubsampling' or something to make it more reasonable. But, anyway, closing this PR.

[NicolasHug]

I agree with you @iaoqian , the current naming might be a bit misleading. Unfortunately we can't really change it because it's public and it would require a deprecation (not sure that it's really worth it). But thanks for sending this PR and for raising this to our attention.
What we can do to avoid such confusion in the future is to clarify in the docstring that the code is intended as-is (i.e. that the Module is just doing subsampling, not max-pooling). Perhaps we can re-purpose your PR for that?

[kkaiwwana]

I understand @NicolasHug. I think we can repurpose this PR by adding some code comments to avoid this issue.

[NicolasHug]

Thank you @iaoqian !