I received a suggestion from mastodon: Cogito
I'm downloading cogito:14b . It also has tools support. Let's see how it'll turn out to be.
I tested a few LLMs locally, asking code and general questions on an M4 Mac Mini. I tried 14B or the closest parameter-sized models since I have 24GB RAM:
Deepseek R1-14B ended up being my favorite model overall.
Besides that, I didn’t have many alternatives with tool support anyway — mostly Llama 3.1-8B and Qwen 2.5-14B. Both are decent, but Llama 3.1 feels slightly better to me at the moment.
If you have any suggestions, I'd love to hear them!
#llm #deepseek #qwen #llama
Hello, thank you for the suggestion. I didn't know about this relay software neither. I believe This has all the features I need, even like cloud backup, and private relay, bundled, and written in #golang .
Also, one of the collaborators already dockerized it already:
https://github.com/bitvora/haven/issues/64#issuecomment-2645375855
https://github.com/aaccioly-open-source/haven
So I can expect a docker hub or ghcr image really soon, or even make myself using github actions. Thank you!
Heya, thank you for suggesting strfry, I didn't know about this software at all, and it looks it's way simpler, and I believe it'll cover my needs since it has whitelist plugin!
https://github.com/hoytech/strfry/blob/master/docs/plugins.md#example-whitelist
Oh and, THANK YOU for damus.io relay and other contributions too. I'm quite new to nostr, but I think it's maintained by you and it's one of the biggest relays so your contribution to the ecosystem so far is definitely invaluable!
Heya, anyone hosting their own personal relay for guaranteed post persistency? What software do you use? I checked, and #nostream could do this, but it's not totally dockerized as far as I can see (no docker images), and building with compose and upgrading would be a hassle if you'd ask me. #asknostr
🇹🇷 MAGA'cı PHP'ci de hiç çekilmiyor ya.
American Rapper Kanye West Launched His First Stream on Twitch and Was Banned After 7 Minutes
In that time, he made a Nazi salute, insulted Jewish people, and made racist and homophobic comments.
Update: Eser Özvataf confirmed this on X platform. It's dependent on "tool use" support of the models. qwen2.5-coder supports it, but deepseek models that I've downloaded did not.
I decided to try running LLMs locally last night, but ran into an issue with agent mode:
I set up LLM Studio and tested Qwen Coder 2.5 14B — didn’t work in agent mode using continue. Switched to Deepseek Coder v2 16B, still no luck. Thought, “Huh, that’s odd.”
Then I installed Ollama, pulled Qwen from there — agent mode worked right away. But with Deepseek, it still said agent mode was unsupported.
I couldn’t figure out the difference — kinda confused.
What are you using locally, and how’s it working for you?
#LLM #LocalLLM #Qwen #DeepSeek #Ollama #AI #AgentMode #OpenSourceAI #MachineLearning #LLMStudio #continue #vscode #programming
New Blog Post: Removing Sensitive Files from Git History: A Step-by-Step Guide
https://arda.pw/posts/remove-sensitive-files-from-git-history/
#git #programming #development
Just a week ago, OpenAI mentioned they were losing / wasting millions because people were saying "thank you" to ChatGPT. Now since they increased the limit, Sam altman reminds people to thank to ChatGPT on X platform instead.
🤡 🤡 🤡 🤡
#chatgpt #samaltman #openai
🫠🫠🫠
Goygoy veya içkili ortamda küfürlü REST ve HTTP kodları anlatma biçimim:
10x = bekle pampa
20x = buyur canım
30x = siktir git
40x = sen sıçtın
50x = ben sıçtım
My little ones are growing so fast!!! #Sgmm
https://video.nostr.build/02504110e02290e2bdb464c0614c5651dc753a6543b43075052cbebdfab079eb.mp4
Denizli mi orası?
I ended up hosting a lightning node over Alby, and paid about $16 equivalent of BTC on lightning network to connect my node to the other nodes. Gladly one of the crptocurrency exchanges that I registered back then had Lightning wallet withdrawals. Well, it's something.