password-manager-resources VS coremltools

Even Apple was so annoyed at this themselves that they actually went for a full open-source open-for-contributions GitHub repository at https://github.com/apple/password-manager-resources to get around these issues.

> Many password managers generate strong, unique passwords for people so that they aren't tempted to create their passwords by hand, which leads to easily guessed and reused passwords. Every time a password manager generates a password that isn't compatible with a website, a person not only has a bad experience but a reason to be tempted to create their password. Compiling password rule quirks helps fewer people run into issues like these while also documenting that a service's password policy is too restrictive for people using password managers, which may incentivize the services to change.

Check out https://github.com/apple/password-manager-resources

Apple has already created the database for this and made it open source: https://github.com/apple/password-manager-resources

Something like this?

Password rules are really all over the place. Based on the sampling available on Apple's password rules database, seems that the majority of sites would accept a 12-character password (although ironically, most websites that restrict the password to be shorter than 12 characters seem to be banks...).

There are still some things in Keychain that feel stupid. For example, Keychain won't merge https://www.google.co.uk and https://www.google.com accounts into one and you can't do it by yourself, and it will even warn about duplicated passwords for these two websites — that's very stupid especially because Apple maintains open database for password managers which solves the problem of alias domains. But that's the most annoying thing for me.

https://github.com/apple/password-manager-resources/blob/mai...

For being "quite obscure", I've at least heard of most of these sites before. Banks with "maxlength: 8", you love to see it.

FWIW, Apple asks users to tell them the password requirements to websites they notice the "Strong Password" feature doesn't work correctly.

The password complexity rule set is open source, you can contribute requirements for specific sites: https://github.com/apple/password-manager-resources

Instead of trying to integrate the whole stack of, say, pytorch, Apple's primary approach has been converting models to work with Apple's stack.

https://github.com/apple/coremltools

Clearly no one is going to be doing training or even fine tuning on Apple hardware at any scale (it competes at the low end, but at scale you invariably will be using nvidia hardware), but once you have a decent model it's a robust way of using it on Apple devices.

There is one guy who was able to run it on iOS. See this thread for more information. Basically, the idea is to convert torch models to CoreMl. Only the CLIP tokenizer's implementation is currently missing. I guess this guy will keep modifications private, but he is trying to optimize model for lower RAM requirements.

Afaik (correct me if I’m wrong) both PyTorch and tensorflow only use the gpu when training and not the neural engine. I think the neural engines can be used for inference if the model is in the CoreML format (https://github.com/apple/coremltools)

CoreML Tools

And Python.

There's literally an Apple provided tool, called [coremltools[(https://github.com/apple/coremltools) to convert many common PyTorch and TensorFlow models to CoreML.