Ciro Santilli OurBigBook.com 新疆改造中心、六四事件、法轮功、郝海东 @cirosantilli

https://t.co/3qttosUjz8 Articles: https://t.co/tK3nhT6FAH Sponsor me: https://t.co/TJ6TaKhfy7 Mastodon: https://t.co/cafDO9wnml @ me if you Tweet about my stuff cirosantilli.com Joined February 2010

Tweets

1K
Followers

2K
Following

13
Likes

2K

Ciro Santilli OurBigBook.com 新疆改造中心、六四事件、法轮功、郝海东 @cirosantilli

a month ago

The following day, the professional spammers with Kenyan IPs (not part of a VPN known to spur.us e.g. spur.us/context/41.90.… ) created 24 SPAM accounts from 2 IPs, now blacklisted. Gotta meet the sweatshop quotas I guess. Sample SPAM: ourbigbook.com/laynemac/hire-…

OurBigBook.com @OurBigBook

a month ago

0 0 0 332 0

0 0 1 309 0

Ciro Santilli OurBigBook.com 新疆改造中心、六四事件、法轮功、郝海东 @cirosantilli

2 months ago

reddit.com/r/selfhosted/c… I did a quick review of LLM inference speeds. A quick rule of thumb for output generation seems to be: 10s of tokens / s on CPU, 100 - 150 on GPU and 1k/2k on Cerebras

0 0 1 471 0

Ciro Santilli OurBigBook.com 新疆改造中心、六四事件、法轮功、郝海东 @cirosantilli

2 months ago

github.com/ollama/ollama/… I managed to produce deterministic LLM output with llama.cpp, the backend of ollama, via --temperature 0 flag, and also run on AMD GPU with their Vulkan backend. These things were very hard/impossible on ollama for silly reasons github.com/ollama/ollama/…