search

ViCCo-Group / thingsvision

Python package for extracting representations from state-of-the-art computer vision models
https://vicco-group.github.io/thingsvision/
MIT License
157 stars 21 forks source link

add condition to allow tensorflow-macos in addition to tensorflow #98

Closed Alxmrphi closed 2 years ago

Alxmrphi commented 2 years ago

Although Lukas said to me in a recent discussion that all contributors use Macs, you must have all solved the tensorflow requirements issue that I and a few others have faced when trying to install THINGSvision on a Macbook. I have to get around this by cloning the repo and manually editing setup.py to allow for tensorflow-macos instead of plain tensorflow and running python setup.py install manually.

This PR adds conditions to both requirements.txt and setup.py to reflect conditional dependency specification (à la PEP508) based on the OS (so the installation doesn't fail if the user has a sufficient version of tensorflow-macos).

This would make it easier for me to keep up to date with the fast-moving pace of this library and hopefully for others who might struggle with the same issues I've faced.

LukasMut commented 2 years ago

Hi Alex! Thanks so much for the contribution and your feedback. I am sorry to hear that this was necessary to get things running on your Mac. It was not necessary for me but might have been for others. I will accept the PR as soon as all checks have passed. :)

Alxmrphi commented 2 years ago

Thanks, Lukas. I'm really happy that this package exists so I thought I'd take your suggestion in the IT & Cognition group and whenever I see something that I think could be better, I'll try to pitch in if there is anything I can do to help :)

LukasMut commented 2 years ago

That's wonderful. I am happy to welcome any new contributors that are willing to help us improve this project. We are actually thinking about releasing a new version of it under a different name some time next year (maybe with a short paper that lists all contributors). I consider the package to be very useful for the community but sometimes it can be difficult to handle in a pretty small team. So, the more we are the merrier. :)