Macosx-15.0-arm64 'is not a supported wheel on this platform' error

,
1

Hi folks!

Which wheel should I use for my macosx-15.0-arm64 laptop? I've tried a bunch of them/debugging but failed.

Thanks!!

2

Welcome to the forum @APieceOfKnife :wave: !

Which Prodigy version are you trying to install and which Python version are you running?

I'm suspecting you're installing pre 1.16.0?
As of 1.16.0 we dropped the support for platform specific wheels and we're distributing cross-platfrom wheels instead.

The last macos-specific wheels (1.15.8) was build for macos-11.0 and Python 3.9 - 3.12. The superior version of the platform should not be a problem but Python version perhaps is? It's important that the Python version you're running matches the python version the wheels were built for.

3

Hi!

Thank you for your info!

The newest one I have is 1.13.3.
The Python version I'm using is 3.13.1, which might be the problem I guess?

Meanwhile my fellow tried to install prodigy 1.12.6 with python 3.10 and it was installed okay. But prodigy stats gave him many compatibility errors - unable to resolved versions and some are pydantic related.

4

Hi @APieceOfKnife ,

In your case that would be the python version, yeah. Would it be possible for you to use Python 3.12 for Prodigy virtual env?

As for the 1.12.6 - I assume they tried to install Prodigy in a fresh virtual env, right? In that case some dependencies are probably not tight enough on 1.12.6. I'll have a look of course but in the meantime the best thing to do would be to upgrade Prodigy to highest version possible.

5

Hi @APieceOfKnife,

With respect to 1.12.6 and Python 3.10: one issue I identified was the numpy version > 2.0.0 which is still allowed in this Prodigy version. The version 2.0.0 of numpy introduced some backward incompatible changes and we pinned it in version 1.15.6 of Prodigy.
Downgrading numpy to 1.26.4 should fix it. I haven't reproduced any Pydantic related issues so I'm assuming it might have come from a Pydantic version installed on the machine unrelated to Prodigy? In an isolated virtual environment, with numpy downgraded, it should be possible to run Prodigy 1.12.6 and Python 3.10.

6

Hi magdaaniol!

Thank you very much for helping this out. The final solution on my side was getting the newest 1.15.8 with python3.12 - which runs perfectly now.

I also tried 1.12.6 and Python 3.10 + downgrade numpy to 1.26.4. It runs!

Thank you again for all your help and hard work.
Have a great day!

7

Awesome! Thanks for reporting back :slight_smile:

1 Like