Skip to content
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

Update ResNet-50 accuracy with Repeated Augmentation #5201

Merged
merged 3 commits into from
Jan 17, 2022

Conversation

tbennun
Copy link
Contributor

@tbennun tbennun commented Jan 16, 2022

cc @datumbox @bjuncek

Updates the accuracy of ResNet-50 on ImageNet to the recipe that uses four Repeated Augmentations (RA) per batch.

Updated recipe: #3995 (comment)

PR is only missing new URL to weights.

Edit:
New weights verified with:

torchrun --nproc_per_node=1 train.py --test-only --weights ResNet50_Weights.ImageNet1K_V2 --model resnet50 -b 1
Acc@1 80.858 Acc@5 95.434

@facebook-github-bot
Copy link

facebook-github-bot commented Jan 16, 2022

💊 CI failures summary and remediations

As of commit 341ca10 (more details on the Dr. CI page):


None of the CI failures appear to be your fault 💚



🚧 1 ongoing upstream failure:

These were probably caused by upstream breakages that are not fixed 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.

Copy link
Contributor

@datumbox datumbox left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@tbennun Very nice, that improves SOTA for ResNet50!

It's awesome to see contributions on the model training side from the community. We plan to improve our process around this to facilitate more contributions. Thanks for your help and patience while we are trying to figure this out.

@datumbox datumbox merged commit f923aeb into pytorch:main Jan 17, 2022
facebook-github-bot pushed a commit that referenced this pull request Jan 19, 2022
Summary:
* Update ResNet-50 accuracy with Repeated Augmentations

* Update the ResNet50 weights URL.

Reviewed By: datumbox, NicolasHug

Differential Revision: D33655256

fbshipit-source-id: 230c8914e4784f84a57f1b89f4f279c34ba8fc3b

Co-authored-by: Vasilis Vryniotis <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants