davidkunz

9 ms·

61.
by davidkunz 3y ago
Out of personal experience, open source LLMs did not yet reach the quality of GPT 3.5, despite multiple claims with dubious benchmarks. That said, they are already useful as of today and can even run on your local machine. I regularly use t
62.
by davidkunz 3y ago
I got confused because I thought it would set a background color. For text it's inverted and hence correct.
63.
by davidkunz 3y ago
Sorry, I misread your answer. You're right. I automatically assumed setting some background color for which `color: light-dark(white, black);` would be right.
64.
by davidkunz 3y ago
But why is the order switched below? Functional notation: light-dark(light-color, dark-color)
65.
by davidkunz 3y ago
I think `black` and `white` should be switched here: color: light-dark(black, white);
66.
by davidkunz 3y ago
A demo video is in the README (I used Mistral-7b in there).
67.
by davidkunz 3y ago
For smaller models, I'm impressed by Mistral-7b or fine-tuned variants like Zephyr. I use it regularly in Neovim[1] for mundane tasks (grammar correction, summaries, ...). I'm curious how Orca 2 performs, downloading it right now.
68.
Helix 23.10: What's New? [video](youtube.com)
1 points by davidkunz 3y ago|0 comments
69.
by davidkunz 3y ago
Here's the plugin: https://github.com/David-Kunz/gen.nvim
70.
by davidkunz 3y ago
Ollama is great, I recently integrated it into my text editor (Neovim), now I have local AI assistance when coding/writing text. https://youtu.be/FIZt7MinpMY?si=fzcmKq5cw2bniH26
71.
Local LLMs in Neovim: gen.nvim [video](youtube.com)
1 points by davidkunz 3y ago|0 comments
72.
by davidkunz 3y ago
> Pages before are topics. What if there's an important detail hidden in a topic?
73.
by davidkunz 3y ago
> Go’s type system allows preventing both issues in a rather elegant way. Proper enums would be elegant, not this.
74.
by davidkunz 3y ago
Aliens discovered: 0 "The World's leading Alien Hunter"
75.
by davidkunz 3y ago
Thank you Bram for giving us software we can love and putting a :smile on our faces.
76.
by davidkunz 3y ago
I like to build my first version in one file with as few abstractions as possible. No helper functions, unless absolutely necessary, just straight-forward code. Building good abstractions later on is a lot easier than starting with wrong on
77.
by davidkunz 3y ago
They should've called it Vicuna v2.
78.
by davidkunz 3y ago
Fantastic language, hoping this will lead to increased adoption!
79.
by davidkunz 3y ago
It's called BlazingMQ but not written in Rust?!
80.
by davidkunz 3y ago
Great work, Stability! Note: It's "Llama 2", not "LLaMA 2", they changed the capitalization.
81.
by davidkunz 3y ago
They used ChatGPT-0301, it can't beat GPT-4.
82.
by davidkunz 3y ago
Me too, the articles are great and I'm excited about the future of Vale.
83.
by davidkunz 3y ago
The "compressed" knowledge is fixed. But reasoning can create new knowledge, that's what humans do all the time.
84.
by davidkunz 3y ago
If the output of prompts can be improved with "think step by step", "tree of thoughts", etc., one can produce high quality training data with the LLM itself which the next iteration of that model can use. Rinse and repea
85.
by davidkunz 3y ago
In JavaScript, primitive types are passed by value, objects are passed by reference. But you won't have data races since JavaScript code is executed in a single thread.
86.
by davidkunz 3y ago
I had a hard time with closures when doing async programming. Is that something one should avoid?
87.
by davidkunz 3y ago
They mixed up million with billion.
88.
by davidkunz 3y ago
> the target size for the new model is 1 trillion parameters. Meanwhile, the target size for the free & public versions of ChatGPT is just 175 million in comparison. That's a 5.7x increase in number of the parameters. Did nobody
89.
by davidkunz 3y ago
Just put "written in Rust" in the title, done.
90.
by davidkunz 3y ago
> our 770M T5 model outperforms the 540B PaLM model using only 80% of available data on a benchmark task. That sounds too good, can someone more knowledgeable comment?

More ›