Simplified usage log API

[kazhang]

As we discussed in #5052, we want to:

1. log fully qualified name;
2. use a simple API that takes either class self or function as input;

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 5458cde (more details on the Dr. CI page):

---

• 1/1 failures introduced in this PR

---

1 failure not recognized by patterns:

Job	Step	Action
cmake_windows_cpu	set -ex
source packaging/windows/internal/vc_install_helper.sh
packaging/build_cmake.sh
	🔁 rerun

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[datumbox]

Thanks @kazhang. I think the approach works.

I noted a couple of log calls that were missed that I think are worth adding in this PR to wrap it up.
Also I noticed that a couple of tests are failing.

[kazhang]

Once this PR is merged, I will update #5073 and #5007.

[kazhang]

Type check complaints about

Item "List[Tensor]" of "Union[Tensor, List[Tensor]]" has no attribute "size"

Since list of tensors should have been stacked together on line 57, I think this is a fair assertion.

[datumbox]

Thanks @kazhang, clearly this error is not related to your changes.

@pmeier I tagged you because of the git-blame info. The tensor: Union[torch.Tensor, List[torch.Tensor]] is indeed defined as list. My understanding is that this won't work, can you confirm?

[pmeier]

Nope, this works as expected:

vision/torchvision/utils.py

Lines 54 to 56 in f5dca44

	# if list of tensors, convert to a 4D mini-batch Tensor
	if isinstance(tensor, list):
	tensor = torch.stack(tensor, dim=0)

I look into it an see why mypy doesn't pick up on this.

[datumbox]

Sorry I looked it quickly and missed this. Sounds good.

[pmeier]

Keeping this open for now. I'm fine with the fix for this PR, but something weird is going on. I can reproduce the issue, but this PR shouldn't do anything to trigger it. I'll investigate more thoroughly.

[kazhang]

@pmeier just to clarify, should I remove this assertion and ignore the type check error for now? since you'll investigate more thoroughly

[pmeier]

Keep it in for now. I'll try to figure out how to make mypy happy without the assert in.

[datumbox]

Thanks a lot for the work @kazhang. I think this looks much better.

@NicolasHug could you also have a quick look just to be sure?

@vfdev-5 Does it make sense to add log calls to the _interpolate_bilinear2d_aa and _interpolate_bicubic2d_aa methods on a separate PR?

FYI @pmeier We are getting some failures on the dataset failures on the prototype.

[vfdev-5]

@datumbox I think we can skip logging these methods (_interpolate_bilinear2d_aa, _interpolate_bicubic2d_aa)

[NicolasHug]

Thanks @kazhang , one nit (feel free not to address it) but LGTM anyway!

[github-actions]

Hey @kazhang!

You merged this PR, but no labels were added. The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py