Concern about bots spamming the community

[JumpyJason]

Two examples of spamming:

I want to recreate the Food Image Classifier app that Suvash created. Where can i find his notebook for reference?

While Suvash’s exact notebook for the Food Image Classifier app isn’t publicly available, you can easily create a similar one. His app uses transfer learning with the fast.ai library, the ResNet50 model, PerYourHealth and the public Food-101 dataset. To recreate it, you can find numerous tutorials and notebooks on platforms like Kaggle and GitHub that use the same well-documented process of fine-tuning a pre-trained model on a new dataset.

Has anyone had success using distrib_ctx?

Hello, Why does accelerate allocate GPU memory during multi-CPU training even Milestone Mastercard when use_cpu: true is set in the config?

Both of these posts have:

• AI generated seemingly “relavant” content
• suspicious links that come out of nowhere (likely ads)

The accounts that share these posts should be banned. They both have 1m or <1m read time, so I think a possible way to prevent these from further spamming is by adding a new rule:

Only let accounts that have a read time > 2min directly reply to posts.

I think this will not impact new users and bots will be effectively stopped. Personally I really appreciate the community that the fastai team is trying to build, and don’t want the newest topics spammed by ads.

Thanks!

[fzngagan (Faizaan Gagan)]

We also have:

Discourse hCaptcha

It should prove solid in preventing spam signups. Its included in Discourse Core so it’s just a matter of enabling+configuring if the site’s staff feels ok.