This week it released a compressed version of Alibaba’s open-source Qwen model, reducing the size of the model from 54GB to 4GB while allowing all 27 billion model parameters to run on an iPhone.
https://www.applemust.com/prism-confirms-apple-is-evaluating-its-llm-compression-tech/
7
Replies
·
0
Boosts
@jonnyevans Hiya. I presume “Prism claims its compressed models use up to 15x memory” is supposed to be "less memory”?
@JasonAnthonyGuy you have correctly identified an error committed by the human curator!
@JasonAnthonyGuy sadly I have been spotting regular errors in these posts. It irritates me so much and I have put myself on a performance review!
@jonnyevans I also usually find my glaring typos after I've pressed publish. Much better than constant proofreading 😆
(I do use a chatbot to try catching them before publishing. It's about 85% successful, which is an improvement over my usual rate.)
@JasonAnthonyGuy I use them all and they still slip through! Often it's when I do a hasty last read and forget an additional check. Other times it's when I'm tired and my eyes don't want to look at the display any more. On the plus side, perhaps errors are the new rock and roll as they show it is written by a person, not a machine! But ChatGPT will probably make a 'create human like' error task now I said that.
@jonnyevans Any of a dozen services will happily sell you that “feature” today…. 🤣🤔😕😭
