Why MMAP in llama.cpp hides true memory usage(twitter.com)

95 points by flurly 3 years ago | 69 comments

Llama.cpp can do 40 tok/s on M2 Max, 0% CPU usage, using all 38 GPU cores(twitter.com)

40 points by samwillis 3 years ago | 9 comments

GPT4 is 8 x 220B params = 1.7T params(twitter.com)

322 points by georgehill 3 years ago | 167 comments

Tweet thread by swyx announcing a Latent Space podcast interview with George Hotz about tiny corp, including claims that GPT-4 is an 8x220B mixture model.

Ohio cryptic lineage update(twitter.com)

145 points by ano-ther 3 years ago | 102 comments

In 2017, developers competed for who could create worst volume control interface(twitter.com)

103 points by rmason 3 years ago | 22 comments

My total comp for creating GitHub Copilot(twitter.com)

108 points by FlamingMoe 3 years ago | 60 comments