Spaces:

huggingchat

/

chat-ui

like 1.09k

Running

The main reason was mostly low usage, we have to put aside a fair amount of GPUs to keep it available at all time but it was not being used enough to justify it, as it's no longer the newest hottest model 😁 Pinning those resources to Falcon also prevents us from showcasing other cool newer models so it's a trade-off we have to make.

@nsarrazin please read and reply even if it's a no. 😢
you can allocate very little resources (the bare minimum) to falcon180b-chat, even it gets overloaded or responds slow, we can still access our hanging chats every now and then. also, there will be a 180B model in the list too (even if it worked only sometimes). or at the very least is there any hope to see it again after a while?
& thank you very much for your quick response! and i sincerely apologize for this and for taking your time.

& some nonsense ig:
another option is limited time availability, like 1 or 2 hours at off hours auto allocation to f180b-c. 4AM or so (regarding timezone there) (could be written there at the list). but i don't think you would even think of this as an option, trying luck ig

Yep I would like to use falcon-180b too... I'm not very active but there's someday u come back and ask something to that big chungus model... There should be a slower instance or only one (if it's not very used)

@maprofi but it's gone now. 180B model running at scale is really expensive. & The traffic wasn't even going there (I'm not experienced in auto scale resources management upon traffic).
A prior notice would've really helped at the very least. Could tidy things up. But let's not dwell on that now. I hope we'll get prior notice in such future scenarios.