-
Notifications
You must be signed in to change notification settings - Fork 6.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Let v2.functional.gaussian_blur
backprop through sigma
parameter
#8450
Comments
Hi @NicolasHug, can I work on this issue? Can you please guide me with a starting point on where can I start looking for? |
Hi Bhavay, I see you've commented on a bunch of issues/PRs on torchvision. At this time, torchvision is a fairly mature project and unfortunately there are no low-hanging fruit PRs that are easy to address for beginners. If you're trying to hone your open source skills, I'd advise contributing into up-and-coming github repos rather than something like torchvision. (You can usethe github search for that) That being said, I'd also encourage you to simply just try contributing and submit a PR instead of asking for guidance of all these posts. For example here in this issue, I feel like the issue is descriptive enough to get someone started (perhaps with some non-trivial initial effort, sure). If you find that it's too advanced for you at this time then that's completely fine, but that probably means you should consider contributing to more beginner friendly repos for now. |
Hi @NicolasHug, thanks for your reply. Actually, I thought that maybe discussing the issue first and then opening a PR would be a good idea but I get your point of try working it our myself first and then opening a draft PR to discuss further. I have some experience in open source and I think I can manage to contribute to torchvision library. In future, I will try to work on issues myself first and then maybe discuss with the team by opening a sample PR. Thanks for the insight. |
the v1 version of
gaussian_blur
allows to backprop through sigma(example taken from #8401)
on CPU and on GPU (after #8426).
However, the v2 version fails with
The support in v1 is sort of undocumented and probably just works out of luck (sigma is typically expected to be a list of floats rather than a tensor). So while it works, it's not 100% clear to me whether this is a feature we absolutely want. I guess we can implement it if it doesn't make the code much more complex or slower.
The text was updated successfully, but these errors were encountered: