Added VicReg - Variance-Invariance-Covariance Regularization for SSL

[dewball345]

This is a reupload(sorry about that). Saw that the original PR was a little hard to find so I created a new updated one.

@owenvallis approved the code but it needs @fchollet's approval to merge into main.

For reference here is the original PR: https://github.com/keras-team/keras-io/pull/847

[dewball345]

Thanks for the PR. I will review shortly. Question: why delete the previous example entirely?

I didn't delete the previous example, at least I don't think so? If you mean the barlow twins notebook, I deleted that because this PR was about VicReg(the barlow twins was accidentally carried on due to a mistake in cloning) - had a separate PR for the barlow twins notebook

[fchollet]

I didn't delete the previous example, at least I don't think so? If you mean the barlow twins notebook, I deleted that because this PR was about VicReg(the barlow twins was accidentally carried on due to a mistake in cloning) - had a separate PR for the barlow twins notebook

The current PR appears to be deleting the Barlow Twins example from the codebase altogether. Please rebase the PR.

Also, please only include the .py file in the PR until the PR is approved (since we would have to regenerate the other files at that time). There should only be one file in the "modified files" tab. Thank you!

[dewball345]

I didn't delete the previous example, at least I don't think so? If you mean the barlow twins notebook, I deleted that because this PR was about VicReg(the barlow twins was accidentally carried on due to a mistake in cloning) - had a separate PR for the barlow twins notebook

The current PR appears to be deleting the Barlow Twins example from the codebase altogether. Please rebase the PR.

Also, please only include the .py file in the PR until the PR is approved (since we would have to regenerated the other files at that time). There should only be one file in the "modified files" tab. Thank you!

Thanks for catching this! Will do!

[sachinprasadhs]

@dewball345 , Could you please update on this PR. Thanks!

[dewball345]

I will see what I can do, let me have a look at this today and get back to you.

On Tue, Aug 8, 2023, 2:36 PM Sachin Prasad @.***> wrote:

@dewball345 https://github.com/dewball345 , Could you please update on this PR. Thanks!

— Reply to this email directly, view it on GitHub https://github.com/keras-team/keras-io/pull/1030#issuecomment-1670341517, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHGJKVDV4Y4XGRLVVSCLCODXUKWPLANCNFSM56SXKUDQ . You are receiving this because you were mentioned.Message ID: @.***>

[dewball345]

@dewball345 , Could you please update on this PR. Thanks!

@sachinprasadhs I really want to thank you for letting me know with regards to updates - i completely overlooked this PR and the change that was requested by @fchollet.

[dewball345]

@sachinprasadhs @fchollet I rebased the PR accordingly and it should be fine now.

[pcoet]

@dewball345 I see the following request from @fchollet:

Also, please only include the .py file in the PR until the PR is approved (since we would have to regenerate the other files at that time). There should only be one file in the "modified files" tab. Thank you!

Can you submit the .py file only, so that @fchollet can review?

[dewball345]

@pcoet I did, the barlow twins files were files that I brought back by rebasing my PR(there were some weird git related things that occurred and the barlow twins notebooks were deleted in this PR so I had to bring them back by a rebase). Those aren't new files, I just brought them back. Only vicreg.py is the file related to vicreg.

[pcoet]

Great, thanks @dewball345! I'll go ahead and assign this to @fchollet.

[dewball345]

Any update on this? @pcoet @fchollet

[pcoet]

@dewball345 Also, I apologize for not requesting these changes earlier. Thanks for your patience!