add positional exception for giou loss

[developer0hye]

vision/torchvision/ops/boxes.py

Lines 301 to 306 in 289fce2

	# degenerate boxes gives inf / nan results
	# so do an early check
	if (boxes1[:, 2:] < boxes1[:, :2]).any():
	raise ValueError("Some of the input boxes1 are invalid.")
	if not (boxes2[:, 2:] >= boxes2[:, :2]).all():
	raise ValueError("Some of the input boxes2 are invalid.")

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 0af1703 (more details on the Dr. CI page):

---

💚 💚 Looks good so far! There are no failures yet. 💚 💚

---

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]

@developer0hye Thanks for the contribution. Unfortunately adding these assertions means a significant slowdown. If you check git blame you will actually see that originally we had assertions and we removed them to improve speed. If you need them then we advise to wrap the loss in your own method and add the assertions there.

[developer0hye]

@datumbox
Thanks for your review!

I also think that it makes this function slow. However, I am still curious about that adding this condition makes whole training process slow.