PR to keras

[keunwoochoi]

Hi, this callback seems quite interesting. Do you plan to PR to the keras repo?

[bckenstler]

You know, I'm actually a novice at open source contributions. I'm not sure exactly how to do that, or whether my callback is interesting enough to merit a contribution. I would love to hear your thoughts!

On Mar 24, 2017 5:08 PM, "Keunwoo Choi" [email protected] wrote:

Hi, this callback seems quite interesting. Do you plan to PR to the keras repo?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/bckenstler/CLR/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/ATJnwTsE8jXDfvzsrReshwex1w63VBnqks5rpFsEgaJpZM4Mo4-n .

[keunwoochoi]

I also hadn't had any experience until a couple of months ago, and actually it was really good experience PR-ing some stuffs. There are unfamiliar aspects that you should know/follow about doing a PR, which I think is quite similar to how people work in the industry these days, and which I was lack of. I'd recommend it to everyone unless s/he is extremely busy.

I think it is interesting enough, got no idea if it's interesting enough for others though. We can always discuss on keras issue channel/twitter. I'll ask @fchollet, too.

[neiljohari]

Hey! I found this repo after reading Smith's paper, and imo this is definitely worthy of a PR. I'm currently in high school so I lack the ability to say whether or not CLR is widely applicable, but from what the paper is stating this seems like a practical tool for any Neural Network project. It's definitely useful to my research project, and I'm sure many others will appreciate this code a lot more if it's in the core of keras. Worst case, your PR is rejected, and that's not a huge deal.

Here's how you can open a PR while meeting all of their guidelines:

• The first thing to look at is to see if there's any special instructions for contributing. A glance at the README.md points us to this contribution guideline file
  • There is a section on Pull Requests, which tells us how to identify which branch to contribute to and what guidelines must be met
• Go ahead and fork fchollet/keras to your own account
• Clone your own fork
• Add your CLR code to the appropriate place (this seems like a good spot)
• Go ahead and make sure you meet all the bullet points in that CONTRIBUTING.md file
  • At a glance, it seems like you're going to be contributing on the keras-contrib branch, and you're going to want to write tests for your new feature by adding to this file

I really hope this all works out. Let me know if you need any help working on the fork/PR, I'd love to help out!

[gewoonrik]

I agree with everyone above. This seems like a nice addition to Keras!

[rivershadowy]

I agree. This is excellent work that should be appreciated and let more people know about.

BTW, it improved my model a lot. :)

[maxfrei750]

@bckenstler Is there any chance, that you will tackle the PR into Keras? If you are not willing to do it yourself, would you allow others to do so?

[YumainOB]

I'm sure your implementation is worthy enough as it is referenced in this pyimagesearch blog post. And as far as I'm concerned it's a nice reference!

EDIT: My bad I just saw that you are in contributors on this keras-contrib callback. Is this callback as reliable as this repos implementation? if yes this issue may be closed to stat this isn't it?