Skip to content

Normalize() transform has undocumented side effect #517

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.

Sign up for GitHub

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

Jump to bottom
Closed
keunhong opened this issue May 26, 2018 · 0 comments
Closed

Normalize() transform has undocumented side effect #517

keunhong opened this issue May 26, 2018 · 0 comments

Comments

@keunhong
Copy link

keunhong commented May 26, 2018 •
edited
Loading

It seems like F.normalize() does all of its operations in-place.

I feel like this should either be very clearly documented or be hidden behind a flag.

This really bit me when I was just sending some batch images to Visdom but my denormalization function was silently denormalizing my batch before doing a forward pass on the network. It was hidden to me until now and surfaces because a copy disappeared when I removed all Variable() wrappers.
@vfdev-5 vfdev-5 mentioned this issue May 26, 2018
Merged
@fmassa fmassa closed this as completed in cf65f39 May 27, 2018
varunagrawal pushed a commit to varunagrawal/vision that referenced this issue Jul 23, 2018
@kohr-h @varunagrawal
Add note on in-place nature of Normalize, closes pytorch#517 (pytorch…
a827f49 
…#520)
rajveerb pushed a commit to rajveerb/vision that referenced this issue Nov 30, 2023
@ahmadki
[SSD] Updated benchmark for for v2.0 (pytorch#517)
8baca74 
* Removed SSD-RN34 benchmark

* [SSD] Updated benchmark for for v2.0 - initial code drop

* [SSD] Added gitignore

* [SSD] fixed transforms and backbone loading

* [SSD] Added inference README

* [SSD] code cleanup, updated bind and train scripts

* [SSD] added eval-print-freq and data-layout options

* [SSD] added support for openimages

* [SSD] improved logging

* [SSD] Added scripts to download openimages

* [SSD] fixed get_database signature

* [SSD] Added a placeholder README

* [SSD] update dataset path

* [SSD] fixed download scripts

* [SSD] Finalized v2 of the benchmark

* OpenImages support
* ADAM optimizer
* New configs
* Bug fixes and performance improvment
* README
* Cleaned backbone and dataset download scripts

* [SSD] README and configs cleanup

* [SSD] fixed small typo in README

* [SSD] Updated README

* Updated the README to resolve some of the open discussions in mlcommons/training#517
* Added a 2-node config equivalent to the single node config
* Minor updates to the launch scripts

* [SSD] Logging compliance with mlperf rules

* [SSD] dataset compliance

* [SSD] fixed the extra brackets in logging

* [SSD] logged tensors initalization

* [SSD] logging fixes
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@keunhong