However, I wonder about the use of this. Typically, you'd want to use a GPU for inference, and programming those in C is quite a challenge. The educational value of learning LLMs in C/C++ seems a lot lower than in Python. May I ask why you're interested in this specifically?
Also, why ask this here, and not simply use Google or an LLM to find the information? It comes off as a bit lazy, especially given that you provide no extra context.
> Also, why ask this here, and not simply use Google or an LLM to find the information? It comes off as a bit lazy
You gave a good answer: why SO the thread? Conversation is the goal here, right? If you’re feeling put out by the question, why bother answering it and then complaining that they asked it? Just move along.
> Also, why ask this here, and not simply use Google or an LLM to find the information? It comes off as a bit lazy
- why answer?
- Google is a fucking mess, no thanks, quicker to ask a person who knows the answer already
- llms don't always give you a decent answer, and the time I spend back and forth would be greater than just asking a human
Why c? Because I want to hand craft a minimal llm without calling in a bunch of libraries that do the actual LLM part for me... I want to know the actual maths going into it. Plus I don't know python. By the looks of it, you can code an llm ina few lines of correctly chosen python; to me this is very unenlightening and I might as well have just ran ollama
> AI by Hand is the research publication of By Hand Research, founded by Prof. Tom Yeh. By Hand Research studies model interpretability and explainability at the math and algorithm level. Subscribers receive free new articles and join live seminars. Members get access to our full research library.
Yes. Convince me why your site is worth my attention, don't try to force it through a dark pattern and then try to justify why I should waste my time figuring it out.
i created something similar a while back. Inspired by micrograd for showing the connection between math/calculus and code, then building along the way to a full NumPy deep learning library that I pretrained GPT-2 124M model with it. One way to learn is to trace through the PRs merged to the repo in chronological order.
It started mostly as an exercise to make sure I actually understood forward propagation and backpropagation rather than just using an existing framework.
Similarly, I’ve been working through Raschka’s Build a Large Language Model (From Scratch) https://sebastianraschka.com/llms-from-scratch/ and it’s been well worth it (particularly paired with his explainer videos)
Train your own LLM - https://github.com/angelos-p/llm-from-scratch
HN Discussion - https://news.ycombinator.com/item?id=48017948