Apple, are you trying to bankrupt us?
At ~500 MB per average model download, 90k hits (Apple's IP addresses are the 17. block) translate to ~45TB of downstream bandwidth… which starts being costly for us.
Your new ML platform Overton (arxiv.org/abs/1909.05372) ⤵️
Conversation
inspired by looks REALLY COOL but maybe at this scale you could add a model caching layer on top 🙂
Hat/tip 🤣🤗
1
2
36
Update for the record: It turns out, it was *not* Overton that was downloading all those models (h/t ) but another project at Apple.
Still curious which one, if you have any insight, feel free to reach out ⚡️
1
11
That's what we ended up doing (temporarily?), but sounds harsh for the other Apple employees :)
1
15
Show replies
Are you going to download a few TBs more? 🤣
You can find the models in this repo: github.com/huggingface/py
1
4
Show replies
Show replies
Haha wow, this makes me really glad we moved all our open-source-related model packages to GitHub releases 😅 (And yet we still get requests from *very* large companies along the lines of "Our firewall blocks GH's AWS, could you host your free models somewhere else just for us?")
31






