Mauve 👁💜 @mauve@mastodon.mauve.moe

It's amazing how much money companies have managed to make from glorified todolist / blog post viewers when that is such a miniscule part of what computers can do.

@neauoire @j3rn Fuck. This is what the word "cyber" should mean in the vernacular. Cybernetics came out of modelling systems, not tracking todo lists. This would be great for system modelling and simulations.

@neauoire @j3rn Are there any conpanies using it ti ship stuff? :O Also how is it's file/network IO?

@j3rn Siiick. Is your source published for those? Would love to read it.

@neauoire @j3rn very cool. What are folks using this for in the wild? :O Would love to see a large peogram written in it to see how folks decompile problems into it 🥰

@neauoire @j3rn Oh my god. I love how this is a text file on an unsecured connrction with ascii text in the header. This is bound to be gold

@j3rn @neauoire Mobile is the worst part in my experience! I'll take a thousand "python module version mismatch" issues over an xcode upgrade 🤣

Neat! What sort of stuff do you even use it for? I've only really looked into it as a curiocity since my job is mostly shoveling bytes around in weird ways rather than working with data.

@j3rn Jeeze that's so true. The main reason I avoid too many layers is people keep breaking them ans forcing me to update/refactor 🤣

I wanna go the other direcrion and code in prolog or something zany just to see what life is like in that world.

@technobaboo Omg yes. Also the characters and music 😭💜

Watching Hackers with my partner finally. God I love this movie.

@simon Nice, like a file format for the configs so folks could pass them around and track changes in git?

@simon PR: https://github.com/simonw/llm-gpt4all/pull/17

Gonna need to mess with the parameters more another day though. But my gut feeling is we can up the quality of output significantly by turning down the temperature a bit and reducing the top_p to 1 and top_k to 4 like in the replicate.com demo

@simon Looked into this, I think the top_p and top_k are the main differences. The default in gpt4all is way more "loose".

Would a PR that sets different defaults be welcome? Or would you prefer to just have the flags exposed like your llama-cpp example?

https://github.com/simonw/llm-gpt4all/blob/main/llm_gpt4all.py#L112

https://docs.gpt4all.io/gpt4all_python.html#the-generate-method-api

I'll try hardcoding some values and running a generation again to see if it's "better" in the meantime.

@pauldaoust Hobestly might replace it with the default textbox implementation in QT/GNOME next. Syntax highlighting is too many cycles!

Fug, I think I am a cranky old fart type already because I like to write code closer to the metal instead of using all the fancy new tooling.

Maybe I should swallow my pride and install vscode and try copilot or whatever it is kids these days are using.

Also I'm still using nano for all my text editing so I might be imposing a deeper limitation on myself than most minimalists. :P

Kinda wild that after years of people being stuck in zoom calls we still need to constantly be like "Can you hear me now?" or "Oh, you're muted" or folks leaving their mic on while loudly typing.

I am guilty of all these things, FWIW. >:P