We have clarified the documentation, sorry about the confusion! 16GB should be enough but it requires some vLLM cache tweaking that we still need to work on, so we put 24GB to be safe. Other deployment methods and quantized versions can definitely fit on 16GB!
Does no one else find it disturbing to publish the identities of these public employees with their salaries? Even if the data is freely available somewhere, this is high visibility. Would the article have been any less interesting if their names were anonymized and their privacy slightly respected?
I find that publishing the names and salaries of random S.F. security guards is quite an unsettling level of transparency.
A few years ago, the package I received contained a bottle of water instead of the phone I ordered. Amazon promptly refunded me and I was quite pleased, because I was a bit concerned by how I would sound to the customer service representative telling this story
The FP16 7B version runs on my Ubuntu XPS with 32GB memory, ~300ms/token. 13B also works but results aren't really good (the model will loop after a few sentences) so parameters probably need tuning.
So far I'm unable to reliably generate outputs in a different language than English, the model will very quickly start to translate (even if it's not asked) or just switch to English.
Absolutely, I meant that the app is not used to collect the email address since it is already in our system but it is true they have no way of knowing.
Funny coincidence that I have received a notification from the Play store this morning saying that analysis of my app (with super tiny outreach) showed that my Data Safety Form was inaccurate and asking me to fix it asap.
Seems like letting the user enter an email address to recover his password counts as email collection (he cannot subscribe, provide or see his email using the app, hence the initial form I submitted).
> The main reason why they don’t care about serving users is because the user is not the customer at Facebook. The users (you and me) are the product sold to the actual customers.
But many people (okay, me at least) just want to watch one or two videos once a week, so a subscription makes no sense... The article also mentions how absurd un-skippable ads have become. The whole "casual YouTube watching experience" is a real pain now, and I do not know if focusing on hard-core tubers will prove right in the long run.